SlideShare une entreprise Scribd logo
1  sur  56
Télécharger pour lire hors ligne
 
	
  
Lehrstuhl und Institut für Arbeitswissenschaft der RWTH Aachen
Direktor:
Univ.-Prof. Dr.-Ing. Dipl.-Wirt.-Ing. Christopher Marc Schlick
Telefon: 0241 80-99 440
Telefax: 0241 80-92 131
info@iaw.rwth-aachen.de
www.iaw.rwth-aachen.de
Dienstgebäude:
Bergdriesch 27 – D-52062 Aachen
Postanschrift:
D-52056 Aachen
	
  
	
   	
   	
   	
   	
   	
  
	
  
	
  
Master	
  Thesis
Robot-human interaction: Investigation of the
uncanny valley using different designed robot alternatives
Angefertigt von
cand.ing. Alesia St. Ivanova
Matr.-Nr.: 328510
Betreuer: Univ.-Prof. Dr.-Ing. Dipl.-Wirt.-Ing. Christopher M. Schlick
Mitbetreuen der
wissenschaftl. Mitarbeiter: Dipl.-Ing. Christopher Brandl
Aachen, den 22.03.2013
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.2	
  /	
  56	
  
	
  
Contents	
  
	
  
1. Introduction	
  ___________________________________________________________	
  	
  p.3	
  
2. Research	
  ______________________________________________________________	
  	
  p.5	
  
1) Previous	
  studies	
  and	
  service	
  robots	
  
i. Care-­‐o-­‐bot	
  
ii. PAMM	
  
iii. RIBA	
  
iv. Taizo	
  
v. Exoskeletons	
  
2) The	
  Uncanny	
  valley	
  
3) Performance	
  assessment	
  
3. The	
  aim	
  ______________________________________________________________	
  	
  p.23	
  
4. Design	
  concept	
  ________________________________________________________	
  	
  p.23	
  
5. Design	
  project	
  _________________________________________________________	
  	
  p.25	
  
6. Pretest	
  design	
  _________________________________________________________	
  	
  p.33	
  
1) Participant	
  and	
  procedure	
  
2) Visualizations	
  	
  
7. Pretest	
  results	
  and	
  analyses	
  ______________________________________________	
  	
  p.48	
  
8. Conclusions	
  ___________________________________________________________	
  	
  p.53	
  
9. References	
  ____________________________________________________________	
  	
  p.55	
  
	
  
	
  
	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.3	
  /	
  56	
  
	
  
Introduction	
  
	
  
	
  
More	
  than	
  30%	
  of	
  the	
  medical	
  expenses	
  all	
  over	
  the	
  world	
  can	
  be	
  related	
  to	
  elderly	
  people.	
  
And	
  these	
  expenses	
  are	
  increasing	
  drastically	
  nowadays.	
  By	
  the	
  year	
  2030	
  the	
  number	
  of	
  60-­‐
year-­‐old	
   people	
   will	
   have	
   doubled.	
   Wheelchair	
   dependents	
   have	
   difficulty	
   moving	
   from	
   a	
  
seat,	
  to	
  their	
  wheelchair	
  and	
  back	
  without	
  a	
  caregivers	
  help	
  or	
  other	
  lift	
  mechanisms.	
  [1]	
  
	
  
While	
  1	
  in	
  3	
  nurses	
  are	
  expected	
  to	
  develop	
  back	
  injuries	
  while	
  moving	
  and	
  lifting	
  patients	
  
and	
  50%	
  of	
  non-­‐ambulatory	
  patients	
  fall	
  to	
  the	
  floor,	
  all	
  aspects	
  of	
  Service	
  robots	
  (e.g.	
  the	
  
HLPR	
   Chair)	
   provide	
   for	
   independent	
   patient	
   mobility	
   especially	
   on	
   lifting	
   and	
   placing	
  
patients	
  to	
  eliminate	
  or	
  significantly	
  reduce	
  this	
  back	
  injury	
  issue.	
  
With	
  fewer	
  caregivers	
  and	
  more	
  elderly,	
  there	
  is	
  a	
  need	
  for	
  improving	
  these	
  technical	
  aids	
  
for	
  providing	
  independent	
  assistance.	
  [2]	
  It	
  is	
  envisioned	
  that	
  in	
  the	
  near	
  future	
  personal	
  
mobile	
  robots	
  will	
  be	
  assisting	
  people	
  in	
  their	
  daily	
  lives.	
  An	
  essential	
  characteristic	
  shaping	
  
the	
  design	
  of	
  personal	
  robots	
  is	
  the	
  fact	
  that	
  they	
  must	
  be	
  accepted	
  by	
  human	
  users.	
  
In	
   general,	
   it	
   seems	
   natural	
   to	
   assume	
   that	
   the	
   more	
   human	
   looking	
   the	
   robots	
   are,	
   the	
  
more	
  likely	
  they	
  are	
  to	
  provoke	
  the	
  usual	
  responses	
  people	
  show	
  to	
  each	
  other.	
  However,	
  
even	
  subtle	
  flaws	
  in	
  appearance	
  and	
  movement	
  seem	
  strange	
  and	
  eerie	
  in	
  very	
  humanlike	
  
robots.	
  	
  	
  	
  	
  
This	
  paper	
  explores	
  the	
  acceptance	
  of	
  mobile	
  personal	
  service	
  robots,	
  by	
  focusing	
  on	
  the	
  
psychological	
  effects	
  of	
  robot	
  appearance.	
  The	
  level	
  of	
  comfort	
  the	
  robot	
  causes	
  to	
  human	
  
subjects	
  is	
  analyzed	
  according	
  to	
  the	
  effects	
  of	
  robot	
  design.	
  The	
  information	
  gained	
  from	
  
surveys	
  taken	
  by	
  40	
  to	
  60	
  human	
  subjects	
  can	
  be	
  used	
  to	
  obtain	
  a	
  better	
  understanding	
  of	
  
what	
  characteristics	
  make	
  up	
  personal	
  robot	
  appearances	
  that	
  are	
  most	
  acceptable	
  to	
  the	
  
human	
   users.	
   The	
   results	
   gained	
   from	
   this	
   study	
   should	
   yield	
   useful	
   insights	
   into	
   how	
   to	
  
calibrate	
  robot	
  appearance	
  so	
  that	
  users	
  of	
  service	
  robots	
  in	
  future	
  will	
  be	
  less	
  disaffected	
  
due	
  to	
  design	
  feature	
  limitations	
  which	
  do	
  not	
  meet	
  their	
  initial	
  expectations.	
  	
  
	
  
This	
   study	
   includes	
   a	
   research	
   about	
   service	
   robot	
   design,	
   users’	
   preferences	
   and	
   the	
  
influence	
   of	
   robot	
   appearance.	
   An	
   overview	
   of	
   the	
   Mori’s	
   “Uncanny	
   valley”	
   hypothesis	
   is	
  
introduced,	
   as	
   well.	
   Furthermore,	
   design	
   alternatives	
   development	
   is	
   presented.	
   Finally,	
  
pretest	
  results	
  and	
  conclusion	
  for	
  further	
  studies	
  are	
  introduced.	
  	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.4	
  /	
  56	
  
	
  
Problem	
  definition	
  
	
  
The	
   empirical	
   studies	
   investigating	
   the	
   importance	
   of	
   the	
   “uncanny	
   valley”	
   all	
   have	
   in	
  
common	
  the	
  low	
  number	
  of	
  factor	
  levels	
  of	
  robot	
  appearances	
  or	
  the	
  equidistance	
  in	
  the	
  
intervals	
  between	
  them.	
  
For	
  example	
  in	
  [3]	
  they	
  used	
  only	
  three	
  “Peoplebot”	
  robot	
  versions	
  in	
  their	
  video-­‐based	
  HRI	
  
trials	
  to	
  support	
  a	
  portion	
  of	
  the	
  left	
  hand	
  side	
  of	
  Mori’s	
  theoretically	
  proposed	
  “uncanny	
  
valley”	
  diagram.	
  	
  
In	
  [4]	
  only	
  highly	
  realistic	
  robots	
  were	
  used	
  to	
  empirically	
  test	
  the	
  right	
  hand	
  side	
  of	
  the	
  
hypothesis	
   by	
   the	
   following	
   statements:	
   a)	
   highly	
   realistic	
   robots	
   are	
   liked	
   less	
   than	
   real	
  
humans	
  and	
  b)	
  the	
  highly	
  realistic	
  robot’s	
  movement	
  decreases	
  its	
  likeability.	
  
According	
  to	
  MacDorman’s	
  analyses	
  on	
  Mori’s	
  theory	
  for	
  the	
  uncanny	
  valley	
  it	
  is	
  “possible	
  to	
  
produce	
  a	
  safe	
  familiarity	
  by	
  a	
  non-­‐humanlike	
  design”,	
  concerning	
  the	
  robot	
  appearance	
  and	
  
its	
  acceptance.	
  
More	
  prolonged	
  experiments,	
  using	
  finer	
  gradations	
  of	
  robot	
  appearances	
  and	
  behavior	
  are	
  
required	
  in	
  order	
  to	
  give	
  more	
  data	
  sample	
  points	
  from	
  the	
  diagram	
  and	
  to	
  provide	
  more	
  
extensive	
  evidence,	
  which	
  can	
  then	
  be	
  used	
  to	
  refine	
  the	
  parameters	
  which	
  define	
  human	
  
perception	
  of	
  robot	
  appearance	
  and	
  how	
  these	
  can	
  be	
  applied	
  to	
  developing	
  principles	
  for	
  
robot	
  aesthetics	
  in	
  different	
  user	
  environments.	
  
	
  
	
  
	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.5	
  /	
  56	
  
	
  
Research	
  
	
  
In	
   this	
   section	
   several	
   service	
   robots	
   are	
   presented,	
   some	
   of	
   them	
   already	
   in	
   use,	
   while	
  
others	
   still	
   in	
   a	
   development	
   phase.	
   	
   Also	
   experiments	
   testing	
   the	
   social	
   acceptance	
   of	
  
robots	
   with	
   varied	
   human-­‐likeness,	
   the	
   importance	
   of	
   tentative	
   acceptance	
   properties	
  
together	
  with	
  the	
  results	
  from	
  these	
  studies,	
  are	
  described.	
  	
  
Some	
   of	
   the	
   researched	
   studies	
   explore	
   the	
   interactions	
   between	
   humans	
   and	
   mobile	
  
personal	
  robots,	
  by	
  focusing	
  on	
  the	
  psychological	
  effects	
  of	
  robot	
  behavior	
  patterns	
  during	
  
task	
  performance.	
  The	
  aim	
  of	
  the	
  research	
  is	
  to	
  gain	
  some	
  guidelines	
  for	
  robot	
  design	
  from	
  
past	
  studies	
  in	
  this	
  area.	
  	
  
	
  
1. Care-­‐O-­‐bot	
  	
  
	
  
Care-­‐O-­‐bot	
  was	
  designed	
  and	
  implemented	
  by	
  Fraunhofer	
  IPA,	
  Stuttgart.	
  The	
  Care-­‐O-­‐bot	
  is	
  a	
  
mobile	
  service	
  robot,	
  which	
  has	
  the	
  capability	
  to	
  perform	
  fetch	
  and	
  carry	
  and	
  various	
  other	
  
supporting	
   tasks	
   in	
   home	
   environments.	
   Main	
   emphasis	
   is	
   laid	
   on	
   integrating	
  
communicational	
  and	
  social	
  features,	
  like	
  video	
  telephone,	
  automatic	
  emergency	
  calls	
  and	
  
other	
  interactive	
  communication	
  (Figure	
  1).	
  
	
  
	
  
Figure	
  1.	
  Care-­‐o-­‐bot®3	
  in	
  action	
  [1]
	
  
Care-­‐O-­‐bot®	
  3 was	
  released	
  relatively	
  soon	
  in	
  2008,	
  after	
  over	
  10	
  years	
  of	
  development,	
  and	
  
excels	
  in	
  its	
  user-­‐interaction	
  oriented	
  design.	
  Nevertheless,	
  it	
  is	
  equipped	
  with	
  leading	
  edge	
  
technology,	
  which	
  is	
  highly	
  integrated	
  into	
  a	
  very	
  compact	
  form.	
  This	
  convergence	
  of	
  design	
  
and	
  technology	
  accounts	
  for	
  the	
  idea	
  of	
  Care-­‐O-­‐bot®	
  3	
  being	
  a	
  product	
  vision	
  for	
  a	
  robot	
  
butler,	
  combining	
  technological	
  aspects	
  with	
  a	
  compact	
  and	
  user	
  friendly	
  design.	
  
The	
  primary	
  interface	
  between	
  Care-­‐O-­‐bot®	
  3 and	
  the	
  user	
  consists	
  of	
  a	
  tray	
  attached	
  to	
  the	
  
front	
  of	
  the	
  robot,	
  which	
  carries	
  objects	
  for	
  exchange	
  between	
  the	
  human	
  and	
  the	
  robot.	
  
The	
   tray	
   includes	
   a	
   touch	
   screen	
   and	
   retracts	
   automatically	
   when	
   not	
   in	
   use.	
   The	
   basic	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.6	
  /	
  56	
  
	
  
concept	
  developed	
  was	
  to	
  define	
  two	
  sides	
  of	
  the	
  robot:	
  “Working	
  side”	
  and	
  “Serving	
  side”.	
  
The	
  concept	
  behind	
  using	
  the	
  tray	
  to	
  interact	
  with	
  the	
  user	
  is	
  to	
  reduce	
  possible	
  users’	
  fears	
  
of	
  mechanical	
  parts	
  by	
  having	
  smooth	
  surfaces	
  and	
  a	
  likable	
  appearance	
  (Figure	
  1.a).	
  On	
  the	
  
technical	
  side,	
  it	
  is	
  much	
  easier	
  to	
  ensure	
  collision	
  free	
  interaction	
  with	
  the	
  static	
  tray	
  than	
  
with	
  a	
  robotic	
  arm	
  moving	
  freely	
  in	
  3-­‐D-­‐space.	
  Using	
  these	
  described	
  interaction	
  concept,	
  
Care-­‐O-­‐bot®3	
   enables	
   the	
   safe	
   executing	
   of	
   fetch	
   and	
   carry	
   tasks	
   and	
   thus	
   provides	
   the	
  
potential	
  to	
  operate	
  a	
  mobile,	
  manipulating	
  robot	
  safely	
  in	
  public	
  environments.	
  	
  
Care-­‐O-­‐bot®	
  3	
  is	
  driven	
  by	
  four	
  wheels.	
  Each	
  wheel’s	
  orientation	
  and	
  rotational	
  speed	
  can	
  be	
  
set	
  individually.	
  The	
  wheeled	
  drive	
  was	
  preferred	
  to	
  legged	
  locomotion	
  because	
  of	
  safety	
  (no	
  
risk	
  of	
  falling)	
  and	
  stability	
  during	
  manipulation.	
  [5]	
  
	
  
	
  
Figure	
  1.a).	
  Care-­‐o-­‐bot®3	
  “two-­‐sides”	
  concept	
  [5]	
  
	
  
	
  
2. PAMM	
  	
  
	
  
PAMM	
   (Personal	
   Aid	
   for	
   Mobility	
   and	
   Monitoring)	
   is	
   intended	
   to	
   assist	
   the	
   elderly	
   living	
  
independently	
   or	
   in	
   senior	
   Assisted	
   Living	
   Facilities.	
   It	
   provides	
   physical	
   support	
   and	
  
guidance,	
  and	
  it	
  monitors	
  the	
  user's	
  basic	
  vital	
  signs.	
  
	
  
Figure	
  2	
  summarizes	
  the	
  PAMM	
  Concept	
  design	
  [6].	
  	
  
	
  
For	
   models	
   released	
   in	
   the	
   early	
   21st
	
   century	
   their	
   appearance	
   was	
   closely	
   to	
   machine-­‐
looking,	
  which	
  is	
  very	
  simplistic	
  and	
  does	
  not	
  intend	
  to	
  impress	
  with	
  any	
  modern	
  or	
  futuristic	
  
body	
  (Figure.2.a),	
  while	
  the	
  latest	
  Walking	
  aid	
  devices	
  impress	
  with	
  smoother	
  shapes	
  and	
  
nice	
  look,	
  associated	
  with	
  a	
  personal	
  vehicle	
  (Figure	
  2.b).	
  Their	
  newer	
  technology	
  of	
  driving	
  
allows	
  using	
  only	
  2	
  wheels	
  instead	
  of	
  3	
  or	
  4.	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.7	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  2.	
  PAMM	
  System	
  Concept	
  
	
  
	
  
	
   	
  
Figure	
  2.a).	
  PAMM	
  devices	
  from	
  2000’s	
  [7]	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.8	
  /	
  56	
  
	
  
	
   	
  
Figure	
  2.b).	
  PAMM	
  devices	
  from	
  Toyota,	
  2008	
  [8]	
  
	
  
	
  
	
  
3. RIBA	
  	
  
RIBA	
  (Robot	
  for	
  Interactive	
  Body	
  Assistance)	
  is	
  said	
  to	
  be	
  the	
  first	
  robot	
  that	
  can	
  lift	
  up	
  or	
  set	
  
down	
  a	
  real	
  human	
  (up	
  to	
  61kg/134lbs)	
  from	
  or	
  to	
  a	
  bed	
  or	
  wheelchair.	
  RIBA	
  does	
  this	
  using	
  
a	
   combination	
   of	
   its	
   very	
   strong	
   human-­‐like	
   arms	
   and	
   by	
   novel	
   tactile	
   guidance	
   methods	
  
using	
   high-­‐accuracy	
   tactile	
   sensors.	
   RIBA	
   was	
   developed	
   by	
   integrating	
   RIKEN's	
   control,	
  
sensor,	
  and	
  information	
  processing	
  and	
  TRI's	
  material	
  and	
  structural	
  design	
  technologies.	
  
	
  
It	
  might	
  look	
  like	
  a	
  cross	
  between	
  a	
  snowman	
  and	
  a	
  badly-­‐designed	
  toy	
  polar	
  bear,	
  but	
  the	
  
nursing	
  fraternity	
  should	
  appreciate	
  this	
  robot	
  that	
  can	
  lift	
  patients	
  (Figure.3)	
  in	
  and	
  out	
  of	
  
beds	
   and	
   wheelchairs	
   on	
   command,	
   while	
   at	
   the	
   same	
   time	
   saving	
   nurses’	
   backs	
   and	
  
improving	
  patient	
  care	
  and	
  safety.	
  [9]	
  
The	
  robot’s	
  body	
  is	
  covered	
  with	
  soft	
  materials	
  and	
  the	
  elbow	
  and	
  waist	
  joints	
  are	
  isolated,	
  
making	
   RIBA	
   safe	
   for	
   physical	
   interactions	
   with	
   humans.	
   This	
   softness	
   also	
   contributes	
   to	
  
patient	
  comfort	
  when	
  they	
  are	
  being	
  lifted.	
  A	
  teddy	
  bear	
  shape	
  was	
  deliberately	
  used	
  to	
  put	
  
patients	
  at	
  ease	
  and	
  to	
  give	
  a	
  friendly,	
  non-­‐threatening,	
  appearance.	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.9	
  /	
  56	
  
	
  
	
  
Figure	
  3.	
  Photograph	
  of	
  RIBA-­‐I	
  and	
  RIBA-­‐II	
  lifting	
  a	
  patient	
  
	
  
4. Taizo	
  
The	
   National	
   Institute	
   of	
   Advanced	
   Industrial	
   Science	
   and	
   Technology	
   (AIST)	
   and	
   Ibaraki	
  
Prefectural	
   Health	
   Plaza	
   in	
   Japan	
   are	
   developping	
   ‘Taizo’,	
   a	
   humanoid	
   robot	
   designed	
   to	
  
lead	
  the	
  elderly	
  in	
  physical	
  exercises.Taizo	
  (Figure	
  4.),	
  which	
  is	
  a	
  play	
  on	
  the	
  word	
  “taisou”	
  
meaning	
  “calisthenics”,	
  stands	
  72cm	
  tall	
  and	
  is	
  dressed	
  in	
  a	
  velvety	
  space	
  suit.	
  He	
  sports	
  a	
  
clown-­‐like	
  grin	
  that	
  is	
  supposed	
  to	
  look	
  silly	
  to	
  put	
  the	
  older	
  generation,	
  who	
  are	
  often	
  a	
  
little	
  frightened	
  by	
  new	
  technology,	
  at	
  ease.	
  [10]	
  	
  
	
  	
  
Figure	
  4.	
  Taizo	
  	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.10	
  /	
  56	
  
	
  
	
  
5. Exoskeletons	
  
	
  
An	
  exoskeleton	
  is	
  a	
  distinctive	
  kind	
  of	
  robot	
  to	
  be	
  worn	
  as	
  an	
  overall,	
  effectively	
  supporting	
  
or,	
   in	
   some	
   cases	
   substituting	
   for,	
   the	
   user’s	
   own	
   movements.	
   The	
   development	
   of	
  
exoskeletons	
   can	
   lead	
   to	
   important	
   changes	
   in	
   the	
   rehabilitation	
   of	
   disabled	
   people	
   by	
  
introducing	
   an	
   alternative	
   to	
   wheelchairs.	
   Exoskeletons	
   can	
   be	
   an	
   efficient	
   tool	
   in	
   the	
  
restoration	
  of	
  upper	
  limb	
  functions,	
  and	
  they	
  can	
  support	
  therapists	
  and	
  caregivers	
  in	
  tasks	
  
that	
  require	
  major	
  physical	
  effort.	
  The	
  functionality	
  of	
  exoskeleton	
  can	
  easily	
  be	
  extended	
  by	
  
a	
  “disabled	
  person	
  integrated	
  IT	
  environment”,	
  described	
  by	
  authors.	
  Exoskeletons	
  can	
  also	
  
be	
  adapted	
  to	
  the	
  needs	
  of	
  severely	
  ill	
  or	
  aged	
  people.	
  
	
  
Exoskeletons	
  can	
  be	
  divided	
  into	
  two	
  categories:	
  those	
  for	
  all	
  four	
  extremities	
  (arms/legs)	
  
and	
   those	
   for	
   the	
   lower	
   extremities	
   only	
   (Figure	
   5.).	
   Exoskeletons	
   are	
   controlled	
   by	
   the	
  
user’s	
  movements	
  and	
  do	
  not	
  need	
  any	
  external	
  control	
  terminal	
  (with	
  the	
  exception	
  of	
  a	
  
service	
   terminal).	
   	
   The	
   main	
   parts	
   are:	
   the	
   frame;	
   the	
   power	
   system,	
   including	
   engines,	
  
actuators	
  and	
  batteries;	
  and	
  the	
  control	
  system	
  with	
  sensors.	
  
	
  
Figure	
  5.	
  Examples	
  of	
  medical	
  exoskeletons:	
  	
  
A)	
  HAL5	
  –	
  version	
  for	
  four	
  extremities;	
  B)	
  ReWalk	
  
	
  
	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.11	
  /	
  56	
  
	
  
	
  
Figure	
  6.	
  Robot	
  suits	
  HAL-­‐5	
  
The	
   robotics	
   geeks	
   at	
   Honda	
   have	
   developed	
   an	
   exoskeleton	
   that	
   is	
   worn	
   like	
   shoes	
   to	
  
support	
  the	
  body	
  and	
  protect	
  the	
  joints,	
  something	
  the	
  automaker	
  says	
  could	
  reduce	
  injuries	
  
on	
  assembly	
  lines	
  but	
  also	
  might	
  help	
  the	
  elderly	
  get	
  around	
  more	
  easily.	
  	
  	
  
The	
  device	
  resembles	
  a	
  bicycle	
  seat	
  joined	
  to	
  a	
  pair	
  of	
  shoes	
  and	
  fits	
  between	
  the	
  legs	
  to	
  
help	
  the	
  user	
  walk,	
  crouch	
  and	
  stand	
  without	
  excessive	
  stress	
  on	
  the	
  hips,	
  knees	
  and	
  ankles.	
  
Honda	
   is	
   testing	
   the	
   "walking	
   assist	
   device"	
   (Figure	
   7.b)	
   at	
   a	
   vehicle	
   assembly	
   line	
   in	
  
Sayama,	
  Japan,	
  and	
  says	
  robo-­‐legs	
  could	
  help	
  anyone	
  who	
  spends	
  a	
  lot	
  of	
  time	
  on	
  their	
  feet.	
  
More	
  than	
  that,	
  it	
  could	
  help	
  the	
  elderly	
  and	
  infirm	
  by	
  making	
  it	
  easier	
  to	
  get	
  around.	
  
In	
  contrast	
  to	
  the	
  complexity	
  of	
  the	
  HAL	
  (Figure	
  6.),	
  the	
  Honda	
  devices’	
  simplicity	
  may	
  be	
  
their	
  strength	
  (Figure	
  7.b.)	
  
For	
  example,	
  U3-­‐X	
  personal	
  mobility	
  prototype	
  with	
  its	
  compact	
  size	
  and	
  one-­‐wheel-­‐drive	
  
personal	
  mobility	
  prototype	
  was	
  designed	
  to	
  be	
  friendly	
  to	
  the	
  user	
  and	
  people	
  around	
  it	
  by	
  
making	
  it	
  easier	
  for	
  the	
  rider	
  to	
  reach	
  the	
  ground	
  from	
  the	
  footrest	
  and	
  placing	
  the	
  rider	
  on	
  
roughly	
  the	
  same	
  eye	
  level	
  as	
  other	
  people	
  or	
  pedestrians.	
  [11]	
  
	
  
Figure	
  7.a.	
  Left	
  to	
  right:	
  U3-­‐X	
  personal	
  mobility	
  prototype,	
  Bodyweight	
  Support	
  Assist Device,	
  Stride	
  
Management	
  Support	
  Assist	
  Device,	
  ASIMO	
  humanoid	
  robot	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.12	
  /	
  56	
  
	
  
Figure	
  7.b.	
  Robo-­‐legs	
  from	
  HONDA	
  [12]	
  
	
  
Tendencies	
  in	
  service	
  robot	
  design	
  
	
  
All	
   of	
   the	
   researched	
   robots	
   above	
   show	
   today’s	
   tendencies	
   in	
   human-­‐robot	
   interaction	
  
design	
  –	
  a	
  combination	
  of	
  high	
  technological	
  execution,	
  smooth	
  shapes	
  and	
  user	
  friendly	
  
interface,	
  materials	
  depend	
  on	
  their	
  purpose	
  and	
  use	
  in	
  daily	
  life.	
  For	
  the	
  current	
  study	
  very	
  
important	
   guidelines	
   can	
   be	
   gained	
   from	
   medical	
   purpose	
   robots	
   such	
   as	
   Riba	
   or	
   Taizo,	
  
because	
  they	
  were	
  designed	
  in	
  order	
  to	
  make	
  people	
  feel	
  comfortable	
  and	
  friendly	
  in	
  their	
  
companion.	
  This	
  would	
  help	
  us	
  to	
  choose	
  a	
  similar	
  approach	
  in	
  designing	
  a	
  head	
  and/or	
  a	
  
face	
  for	
  our	
  experimental	
  robots	
  and	
  investigate	
  this	
  part	
  of	
  the	
  uncanny	
  valley	
  graph	
  which	
  
is	
  more	
  human-­‐like.	
  	
  
The	
  other	
  devices	
  like	
  PAMM,	
  Care-­‐o-­‐bot	
  and	
  exoskeletons	
  are	
  shown	
  especially	
  for	
  their	
  
driving	
  systems	
  –	
  how	
  do	
  they	
  execute	
  movements,	
  tasks	
  and	
  how	
  they	
  approach	
  to	
  users.	
  
Obviously	
  the	
  more	
  human-­‐like	
  is	
  the	
  robot,	
  the	
  more	
  complicated	
  “walking”	
  system	
  is.	
  The	
  
machine-­‐looking,	
   such	
   as	
   PAMM	
   and	
   Care-­‐o-­‐bot	
   use	
   wheels,	
   and	
   even	
   more	
   machine-­‐
looking	
  one	
  would	
  have	
  chains	
  like	
  in	
  a	
  tank.	
  This	
  kind	
  of	
  machine	
  parts	
  make	
  the	
  robots	
  
more	
  maneuverable	
  and	
  smooth	
  in	
  movements,	
  while	
  walking-­‐resembling	
  robots	
  are	
  still	
  in	
  
development	
  for	
  improving	
  realistic	
  movements.	
  	
  
	
  
	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.13	
  /	
  56	
  
	
  
The	
  Uncanny	
  valley	
  
	
  
	
  “Climbing	
  a	
  mountain	
  is	
  an	
  example	
  of	
  a	
  function	
  that	
  does	
  not	
  increase	
  continuously:	
  a	
  person's	
  
altitude	
  does	
  not	
  always	
  increase	
  as	
  the	
  distance	
  from	
  the	
  summit	
  decreases	
  owing	
  to	
  the	
  intervening	
  
hills	
  and	
  valleys.	
  I	
  have	
  noticed	
  that,	
  as	
  robots	
  appear	
  more	
  humanlike,	
  our	
  sense	
  of	
  their	
  familiarity	
  
increases	
  until	
  we	
  come	
  to	
  a	
  valley.	
  I	
  call	
  this	
  relation	
  the	
  “uncanny	
  valley”.”	
  
Masahiro	
  Mori	
  
	
  
Recently	
  prosthetic	
  hands	
  have	
  improved	
  greatly,	
  and	
  we	
  cannot	
  distinguish	
  them	
  from	
  real	
  
hands	
   at	
   a	
   glance.	
   Some	
   prosthetic	
   hands	
   attempt	
   to	
   simulate	
   veins,	
   muscles,	
   tendons,	
  
finger	
  nails,	
  and	
  finger	
  prints,	
  and	
  their	
  color	
  resembles	
  human	
  pigmentation.	
  But	
  this	
  kind	
  
of	
  prosthetic	
  hand	
  is	
  too	
  real	
  and	
  when	
  one	
  notices	
  it	
  is	
  prosthetic,	
  they	
  have	
  a	
  sense	
  of	
  
strangeness.	
  So	
  if	
  we	
  shake	
  the	
  hand,	
  we	
  are	
  surprised	
  by	
  the	
  lack	
  of	
  soft	
  tissue	
  and	
  cold	
  
temperature.	
   In	
   this	
   case,	
   there	
   is	
   no	
   longer	
   a	
   sense	
   of	
   familiarity.	
   It	
   is	
   uncanny.	
   In	
  
mathematical	
   terms,	
   strangeness	
   can	
   be	
   represented	
   by	
   negative	
   familiarity,	
   so	
   the	
  
prosthetic	
   hand	
   is	
   at	
   the	
   bottom	
   of	
   the	
   valley.	
   So	
   in	
   this	
   case,	
   the	
   appearance	
   is	
   quite	
  
human	
  like,	
  but	
  the	
  familiarity	
  is	
  negative.	
  This	
  is	
  the	
  uncanny	
  valley.	
  
	
  
	
  
Figure	
  8.	
  The	
  Uncanny	
  Valley	
  graph	
  
	
  
As	
  a	
  robot	
  designer,	
  Mori	
  graphed	
  what	
  he	
  saw	
  as	
  the	
  relation	
  between	
  human	
  likeness	
  and	
  
perceived	
  familiarity:	
  familiarity	
  increases	
  with	
  human	
  likeness	
  until	
  a	
  point	
  is	
  reached	
  at	
  
which	
   deviations	
   from	
   human	
   appearance	
   and	
   behavior	
   create	
   unnerving	
   effect.	
   This	
   he	
  
called	
  the	
  uncanny	
  valley.	
  According	
  to	
  Mori,	
  movement	
  amplifies	
  the	
  effect	
  (Figure	
  8.)	
  
	
  
In	
  the	
  World	
  Expo	
  held	
  in	
  Osaka	
  2009	
  were	
  displayed	
  robots	
  with	
  a	
  very	
  elaborate	
  design.	
  
For	
   example,	
   one	
   robot	
   has	
   29	
   artificial	
   muscles	
   in	
   the	
   face	
   to	
   make	
   humanlike	
   facial	
  
expressions.	
  According	
  to	
  the	
  designer,	
  laughing	
  is	
  a	
  kind	
  of	
  sequence	
  of	
  face	
  distortions,	
  
and	
  the	
  distortion	
  speed	
  is	
  an	
  important	
  factor.	
  If	
  the	
  speed	
  is	
  cut	
  in	
  half,	
  laughing	
  looks	
  
unnatural.	
  This	
  illustrates	
  how	
  slight	
  variations	
  in	
  movement	
  can	
  cause	
  a	
  robot,	
  puppet,	
  or	
  
prosthetic	
  hand	
  to	
  tumble	
  down	
  into	
  the	
  uncanny	
  valley.	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.14	
  /	
  56	
  
	
  
The	
  author	
  hopes	
  to	
  design	
  robots	
  or	
  prosthetic	
  hands	
  that	
  will	
  not	
  fall	
  into	
  the	
  uncanny	
  
valley.	
   So	
   he	
   recommends	
   designers	
   to	
   take	
   the	
   first	
   peak	
   as	
   a	
   goal	
   in	
   building	
   robots	
  
rather	
  than	
  the	
  second.	
  Although	
  the	
  second	
  peak	
  is	
  higher,	
  there	
  is	
  a	
  far	
  greater	
  risk	
  of	
  
falling	
  into	
  the	
  uncanny	
  valley.	
  They	
  predict	
  that	
  it	
  is	
  possible	
  to	
  produce	
  a	
  safe	
  familiarity	
  
by	
  a	
  nonhumanlike	
  design.	
  This	
  is	
  a	
  good	
  point	
  to	
  take	
  into	
  a	
  consideration.	
  	
  
	
  
A	
   good	
   example	
   is	
   glasses.	
   Glasses	
   do	
   not	
   resemble	
   the	
   real	
   eyeball,	
   but	
   this	
   design	
   is	
  
adequate	
  and	
  can	
  make	
  the	
  eyes	
  more	
  charming.	
  So	
  designers	
  should	
  follow	
  this	
  principle	
  
when	
  design	
  prosthetic	
  eyes.	
  An	
  elegant	
  prosthetic	
  hand	
  can	
  be	
  created	
  -­‐	
  one	
  that	
  must	
  be	
  
fashionable.	
  Artist	
  who	
  makes	
  statues	
  of	
  Buddha	
  created	
  a	
  model	
  of	
  a	
  human	
  hand	
  that	
  is	
  
made	
   from	
   wood.	
   The	
   fingers	
   bend	
   at	
   their	
   joints.	
   The	
   hand	
   has	
   no	
   finger	
   print,	
   and	
   it	
  
assumes	
  the	
  natural	
  color	
  of	
  wood.	
  But	
  it	
  still	
  looks	
  beautiful	
  and	
  there	
  is	
  no	
  sense	
  of	
  the	
  
uncanny.	
  [13]	
  
Christoph	
  Bartneck,	
  Takayuki	
  Kanda,	
  Hiroshi	
  Ishiguro,	
  and	
  Norihiro	
  Hagita	
  (Members	
  of	
  IEEE)	
  
conducted	
  a	
  study	
  which	
  attempted	
  to	
  empirically	
  test	
  two	
  aspects	
  of	
  Mori’s	
  hypothesis.	
  
First,	
  they	
  were	
  interested	
  in	
  the	
  degree	
  to	
  which	
  highly	
  realistic	
  androids	
  were	
  perceived	
  
differently	
  from	
  a	
  human.	
  The	
  uncanny	
  valley	
  hypothesis	
  predicts	
  that	
  androids	
  would	
  be	
  
perceived	
  as	
  less	
  human-­‐like	
  and	
  less	
  likeable	
  compared	
  to	
  humans.	
  	
  
To	
  test	
  this	
  hypothesis,	
  they	
  used	
  Hiroshi	
  Ishiguro	
  and	
  his	
  robotic	
  copy	
  named	
  “Geminoid	
  HI-­‐
1”	
   (Figure	
   9).	
   Also,	
   they	
   wanted	
   to	
   test	
   whether	
   a	
   more	
   humanlike	
   android	
   would	
   be	
  
perceived	
  as	
  more	
  likeable	
  compared	
  to	
  a	
  less	
  human-­‐like	
  robot.	
  Accordingly,	
  they	
  made	
  a	
  
small	
  alteration	
  to	
  Geminoid	
  HI-­‐1	
  to	
  make	
  it	
  appear	
  less	
  humanlike.	
  
	
  
	
  
Figure	
  9.	
  Hiroshi	
  Ishiguro	
  and	
  his	
  robotic	
  double	
  Geminoid	
  HI-­‐1	
  
	
  
Second,	
  they	
  were	
  interested	
  in	
  the	
  effect	
  of	
  the	
  android’s	
  movement.	
  Mori’s	
  hypothesis	
  
predicts	
  that	
  movement	
  intensifies	
  the	
  users’	
  perception	
  of	
  an	
  android.	
  A	
  moving	
  android	
  
would	
  be	
  perceived	
  differently	
  from	
  an	
  inert	
  android.	
  	
  
	
  
	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.15	
  /	
  56	
  
	
  
In	
  summary,	
  the	
  following	
  three	
  hypotheses	
  were	
  interesting	
  for	
  the	
  study:	
  
	
  
1.	
  Androids	
  that	
  are	
  distinguishable	
  from	
  humans	
  will	
  be	
  liked	
  less	
  than	
  humans.	
  
2.	
  A	
  fully	
  moving	
  android	
  will	
  be	
  liked	
  differently	
  compared	
  to	
  an	
  android	
  that	
  is	
  limited	
  in	
  its	
  movements.	
  
3.	
  Androids	
  with	
  different	
  levels	
  of	
  anthropomorphism	
  will	
  be	
  liked	
  differently.	
  
	
  
They	
   conducted	
   a	
   3	
   (anthropomorphism)	
   x	
   2	
   (movement)	
   experiment	
   in	
   which	
  
anthropomorphism	
   was	
   the	
   within	
   participants	
   factor	
   and	
   movement	
   was	
   the	
   between	
  
participants	
   factor.	
   The	
   anthropomorphism	
   factor	
   had	
   three	
   conditions:	
   masked	
   android,	
  
android,	
  and	
  human.	
  The	
  movement	
  factor	
  had	
  two	
  conditions:	
  full	
  movement	
  and	
  limited	
  
movement.	
  	
  
The	
   participants	
   in	
   this	
   study	
   were	
   19	
   men	
   and	
   13	
   women	
   in	
   their	
   early	
   20’s	
   attending	
  
Japanese	
  universities	
  in	
  the	
  Kansai	
  area.	
  The	
  male	
  and	
  female	
  participants	
  were	
  distributed	
  
approximately	
   even	
   across	
   the	
   experimental	
   conditions.	
   They	
   were	
   not	
   exposed	
   to	
   any	
  
previous	
  study	
  in	
  the	
  laboratory.	
  This	
  study	
  was	
  conducted	
  shortly	
  before	
  the	
  official	
  release	
  
of	
  the	
  Geminoid	
  HI-­‐1	
  and	
  hence	
  they	
  were	
  not	
  exposed	
  to	
  the	
  considerable	
  media	
  exposure	
  
that	
  the	
  Geminoid	
  HI-­‐1	
  received.	
  
The	
  participants	
  were	
  welcomed	
  and	
  then	
  asked	
  to	
  fill	
  in	
  a	
  questionnaire.	
  They	
  were	
  seated	
  
on	
   a	
   chair	
   that	
   was	
   placed	
   one	
   meter	
   away	
   from	
   the	
   android/person.	
   Afterwards	
   the	
  
experimenter	
  introduced	
  them	
  to	
  each	
  other	
  without	
  explicitly	
  labeling	
  Ishiguro	
  as	
  a	
  human	
  
and	
  the	
  androids	
  as	
  robots.	
  
	
  
Conclusions	
  from	
  the	
  study:	
  Against	
  Mori’s	
  prediction,	
  androids	
  that	
  were	
  distinguishable	
  
from	
   humans	
   were	
   not	
   liked	
   less	
   than	
   humans.	
   The	
   results	
   showed	
   that	
   the	
   participants	
  
were	
  able	
  to	
  distinguish	
  between	
  the	
  human	
  stimulus	
  and	
  the	
  android	
  stimuli.	
  The	
  human	
  
was	
  rated	
  as	
  being	
  significantly	
  more	
  human-­‐like	
  compared	
  to	
  the	
  two	
  androids.	
  However,	
  
the	
  ratings	
  for	
  likeability	
  were	
  not	
  significantly	
  different.	
  This	
  result	
  does	
  not	
  support	
  Mori’s	
  
hypothesis.	
  Two	
  possible	
  interpretations	
  could	
  be	
  possible.	
  On	
  the	
  one	
  hand,	
  there	
  really	
  
could	
  be	
  no	
  difference	
  between	
  the	
  likeability	
  of	
  humans	
  and	
  that	
  of	
  androids.	
  On	
  the	
  other	
  
hand,	
  likeability	
  could	
  be	
  a	
  more	
  complex	
  phenomenon.	
  They	
  speculate	
  that	
  the	
  participants	
  
might	
   have	
   used	
   different	
   standards	
   to	
   evaluate	
   the	
   likeability	
   of	
   the	
   human	
   and	
   the	
  
androids.	
  As	
  a	
  robot,	
  the	
  displayed	
  android	
  might	
  have	
  been	
  likeable	
  to	
  the	
  same	
  degree	
  as	
  
the	
  human	
  was	
  likeable	
  as	
  a	
  human;	
  however,	
  the	
  expectations	
  for	
  these	
  two	
  categories	
  
might	
  have	
  been	
  different.	
  
	
  
The	
  results	
  of	
  this	
  study	
  cannot	
  confirm	
  Mori’s	
  hypothesis	
  of	
  the	
  Uncanny	
  Valley.	
  The	
  robots’	
  
movements	
  and	
  their	
  level	
  of	
  anthropomorphism	
  may	
  be	
  complex	
  phenomena	
  that	
  cannot	
  
be	
   reduced	
   to	
   two	
   factors.	
   Movement	
   contains	
   social	
   meanings	
   that	
   may	
   have	
   direct	
  
influence	
  on	
  the	
  likeability	
  of	
  a	
  robot.	
  The	
  robot’s	
  level	
  of	
  anthropomorphism	
  does	
  not	
  only	
  
depend	
   on	
   its	
   appearance	
   but	
   also	
   on	
   its	
   behavior.	
   A	
   mechanical-­‐looking	
   robot	
   with	
  
appropriate	
  social	
  behavior	
  can	
  be	
  anthropomorphized	
  for	
  different	
  reasons	
  than	
  a	
  highly	
  
humanlike	
  android.	
  Again,	
  Mori’s	
  hypothesis	
  appears	
  to	
  be	
  too	
  simplistic.	
  [4]	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.16	
  /	
  56	
  
	
  
	
  
Figure	
  10.	
  Graph	
  showing	
  results	
  of	
  MacDorman	
  and	
  Hiroshi	
  Ishiguro	
  study	
  on	
  the	
  Uncanny	
  valley	
  
	
  
Another	
  study	
  from	
  MacDorman	
  and	
  Ishiguro	
  is	
  presented	
  here.	
  The	
  average	
  ratings	
  of	
  45	
  
Indonesian	
  participants	
  on	
  scales	
  of	
  human	
  likeness,	
  familiarity,	
  and	
  eeriness	
  are	
  presented	
  
for	
  the	
  above	
  figure	
  (Figure	
  10).	
  The	
  images	
  morph	
  from	
  a	
  mechanical-­‐looking	
  humanoid	
  on	
  
the	
  left	
  to	
  an	
  android	
  in	
  the	
  center	
  to	
  a	
  human	
  being	
  on	
  the	
  right.	
  
	
  
For	
  the	
  given	
  images,	
  they	
  reveal	
  an	
  uncanny	
  region,	
  both	
  on	
  the	
  strange-­‐familiar	
  scale	
  and	
  
on	
   the	
   eeriness	
   scale.	
   In	
   the	
   follow-­‐up	
   discussion	
   David	
   Han-­‐son	
   argued	
   that	
   a	
   valley	
   of	
  
eeriness	
  was	
  not	
  inevitable	
  for	
  a	
  specific	
  range	
  of	
  human	
  likeness.	
  He	
  claims	
  that,	
  across	
  the	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.17	
  /	
  56	
  
	
  
spectrum	
   of	
   human	
   likeness,	
   it	
   is	
   possible	
   to	
   design	
   androids	
   that	
   are	
   not	
   uncanny.	
   In	
   a	
  
follow-­‐up	
   experiment	
   in	
   which	
   intermediate	
   images	
   were	
   designed,	
   adapting	
   more	
  
attractive,	
  cartoon-­‐like	
  features,	
  rather	
  than	
  simply	
  morphed,	
  Hanson	
  eliminates	
  the	
  valley	
  
from	
   his	
   results.	
   In	
   addition,	
   Hanson	
   notes	
   that	
   very	
   abstract	
   robots	
   and	
   cosmetically	
  
atypical	
  people	
  can	
  be	
  uncanny,	
  although	
  they	
  are	
  far	
  from	
  the	
  posited	
  region	
  of	
  the	
  valley	
  
in	
  terms	
  of	
  human	
  likeness.	
  	
  
	
  
Do	
  Hanson’s	
  results	
  mean	
  that	
  the	
  uncanny	
  valley	
  does	
  not	
  exist?	
  They	
  may	
  suggest	
  that	
  the	
  
uncanny	
  valley	
  is	
  not	
  inevitable	
  or	
  that	
  designers	
  with	
  finesse	
  can	
  moderate	
  it	
  in	
  situations	
  
that	
   involve	
   still	
   images.	
   Nevertheless,	
   human	
   beings	
   do	
   seem	
   to	
   be	
   highly	
   sensitive	
   to	
  
imperfections	
  in	
  near	
  humanlike	
  robots,	
  both	
  in	
  their	
  looks	
  and	
  movements,	
  which	
  is	
  why	
  
androids	
  are	
  potentially	
  very	
  useful	
  in	
  studying	
  human	
  perception.	
  Furthermore,	
  only	
  limited	
  
conclusions	
  can	
  be	
  drawn	
  from	
  ratings	
  of	
  still	
  images,	
  which	
  are	
  static,	
  modern	
  inventions	
  
appearing	
   after	
   human	
   beings	
   evolved.	
   For	
   instance,	
   cartoon	
   images	
   can	
   be	
   aesthetically	
  
pleasing,	
  but	
  if	
  real	
  people	
  could	
  exist	
  with	
  the	
  same	
  proportions,	
  they	
  would	
  be	
  considered	
  
freaks.	
  There	
  is	
  no	
  way	
  to	
  evaluate	
  whether	
  a	
  still	
  image	
  is	
  responding	
  as	
  predicted,	
  because	
  
they	
  cannot	
  respond	
  at	
  all.	
  [14]	
  
	
  
	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.18	
  /	
  56	
  
	
  
Performance	
  assessment	
  
	
  
	
  
The	
  effectiveness	
  of	
  service	
  robots	
  cannot	
  be	
  assessed	
  only	
  by	
  performance	
  criteria	
  typically	
  
found	
   for	
   industrial	
   robots.	
   The	
   performance	
   criteria	
   of	
   service	
   robots	
   lie	
   within	
   the	
  
satisfaction	
   of	
   their	
   users.	
   Therefore,	
   it	
   is	
   necessary	
   to	
   measure	
   the	
   users’	
   perception	
   of	
  
service	
   robots,	
   since	
   these	
   cannot	
   be	
   measured	
   within	
   the	
   robots	
   themselves.	
   Below	
   are	
  
some	
  of	
  the	
  main	
  criteria	
  used	
  in	
  these	
  measurements.	
  
	
  
Social	
  acceptance	
  
What	
   can	
   be	
   considered	
   to	
   be	
   the	
   important	
   criteria	
   for	
   social	
   acceptance	
   of	
   robots	
   in	
  
homes	
  and	
  health	
  care?	
  It	
  will	
  most	
  likely	
  vary	
  with	
  both	
  application	
  area	
  and	
  culture.	
  One	
  
example	
  of	
  such	
  criteria	
  might	
  be	
  the	
  size	
  of	
  the	
  robot,	
  where	
  several	
  aspects	
  have	
  been	
  
reported.	
   One	
   such	
   consideration	
   states	
   that	
   the	
   robot	
   has	
   to	
   be	
   smaller	
   than	
   human	
   in	
  
order	
  not	
  to	
  "dominate"	
  the	
  human	
  user.	
  This	
  criteria	
  (size)	
  can	
  be	
  developed	
  further	
  in	
  the	
  
context	
   of	
   home	
   care	
   for	
   handicapped	
   people,	
   where	
   it	
   is	
   stated	
   that	
   the	
   robot	
   has	
   to	
  
defend	
  its	
  space	
  though	
  its	
  functionality,	
  since	
  a	
  large	
  appliance	
  will	
  compete	
  with	
  the	
  other	
  
support	
  equipment	
  that	
  is	
  already	
  in	
  the	
  care	
  environment	
  (such	
  as	
  wheel	
  chairs,	
  respiratory	
  
equipment	
  etc.).	
  If	
  the	
  robot	
  size	
  is	
  too	
  big,	
  it	
  will	
  probably	
  simply	
  not	
  be	
  used.	
  
Examples	
   of	
   similar	
   properties	
   that	
   will	
   be	
   important	
   to	
   look	
   at	
   in	
   the	
   social	
   acceptance	
  
perspective	
  are	
  technical	
  issues	
  such	
  as:	
  
• Weight	
  -­‐	
  the	
  impression	
  of	
  moving	
  weight	
  
• Speed	
  -­‐	
  the	
  speed	
  of	
  the	
  robot’s	
  movement	
  
• Agility	
  -­‐	
  the	
  speed	
  of	
  limb	
  movements	
  
• Reliability	
  -­‐	
  the	
  functional	
  stability	
  of	
  the	
  robot	
  
• There	
  are	
  also	
  more	
  emotional	
  and	
  possibly	
  culture	
  dependent	
  issues	
  such	
  as:	
  
• Anthropomorphism	
  -­‐	
  should	
  the	
  robot	
  look	
  like	
  a	
  human	
  or	
  a	
  machine?	
  
• Social	
  behavior	
  -­‐	
  which	
  behavior	
  characteristics	
  are	
  important	
  for	
  the	
  acceptance	
  of	
  the	
  robot?	
  
• Autonomy	
  -­‐	
  should	
  the	
  robot	
  act	
  with	
  or	
  without	
  user	
  involvement?	
  
• Distance	
  -­‐	
  how	
  big	
  should	
  be	
  the	
  minimum	
  distance	
  between	
  human	
  and	
  robot	
  (something	
  
which	
  clearly	
  is	
  culturally	
  dependent)?	
  
• Safety	
  and	
  security	
  -­‐	
  how	
  safe	
  does	
  a	
  person	
  feel	
  in	
  relation	
  to	
  the	
  robot?	
  
• Reliability	
  -­‐	
  can	
  you	
  trust	
  the	
  robot	
  to	
  perform	
  the	
  tasks	
  and	
  do	
  this	
  in	
  the	
  right	
  way?	
  
	
  
Taken	
  together,	
  these	
  variables	
  will	
  affect	
  the	
  user	
  in	
  one	
  way	
  or	
  another.	
  It	
  is	
  therefore	
  
necessary	
   for	
   the	
   developers	
   and	
   designers	
   to	
   find	
   out	
   which	
   of	
   these	
   aspects	
   that	
   are	
  
important	
  to	
  consider,	
  and	
  second,	
  how	
  the	
  properties	
  affect	
  the	
  social	
  acceptance	
  of	
  the	
  
robot.	
  Below	
  some	
  of	
  them	
  are	
  described	
  in	
  more	
  details.	
  
	
  
Anthropophormism	
  
Anthropomorphism	
   refers	
   to	
   the	
   attribution	
   of	
   a	
   human	
   form,	
   human	
   characteristics,	
   or	
  
human	
  behavior	
  to	
  nonhuman	
  things	
  such	
  as	
  robots,	
  computers,	
  and	
  animals.	
  	
  
	
  
Even	
  if	
  it	
  is	
  not	
  the	
  intention	
  of	
  the	
  design	
  of	
  a	
  certain	
  robot	
  to	
  be	
  as	
  humanlike	
  as	
  possible,	
  
it	
   still	
   remains	
   important	
   to	
   match	
   the	
   appearance	
   of	
   the	
   robot	
   with	
   its	
   abilities.	
   A	
   too	
  
anthropomorphic	
  appearance	
  can	
  evoke	
  expectations	
  that	
  the	
  robot	
  might	
  not	
  be	
  able	
  to	
  
fulfill.	
  If,	
  for	
  example,	
  the	
  robot	
  has	
  a	
  human-­‐shaped	
  face	
  then	
  the	
  naive	
  user	
  will	
  expect	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.19	
  /	
  56	
  
	
  
that	
  the	
  robot	
  is	
  able	
  to	
  listen	
  and	
  to	
  talk.	
  To	
  prevent	
  disappointment	
  it	
  is	
  necessary	
  for	
  all	
  
developers	
  to	
  pay	
  close	
  attention	
  to	
  the	
  anthropomorphism	
  level	
  of	
  their	
  robots.	
  
	
  
Anthropomorphism	
  is	
  a	
  constant	
  pattern	
  in	
  human	
  cognition	
  and	
  the	
  interaction	
  of	
  a	
  human	
  
with	
  a	
  robot	
  (or	
  any	
  kind	
  of	
  machine)	
  cannot	
  completely	
  avoid	
  it.	
  According	
  to	
  Mori,	
  the	
  so-­‐
called	
  uncanny	
  valley	
  would	
  suggest	
  to	
  either	
  stay	
  in	
  the	
  area	
  of	
  very	
  non-­‐human,	
  toy-­‐like	
  
robots,	
  or	
  to	
  create	
  a	
  robot	
  that	
  appears	
  to	
  be	
  almost	
  perfectly	
  human-­‐like,	
  because	
  a	
  robot	
  
in	
  between	
  may	
  provoke	
  rather	
  fearful	
  responses.	
  [15].	
  Human-­‐like	
  appearance	
  is	
  likely	
  to	
  
trigger	
  expectations	
  that	
  go	
  beyond	
  the	
  capabilities	
  of	
  a	
  machine.	
  But	
  being	
  humanoid	
  in	
  
appearance	
  does	
  hardly	
  suffice	
  to	
  meet	
  the	
  expectancy	
  of	
  humanlike	
  reactions.	
  [16]	
  
	
  
Likeability	
  
There	
   is	
   a	
   growing	
   body	
   of	
   research	
   which	
   indicates	
   that	
   people	
   often	
   make	
   important	
  
judgments	
  within	
  seconds	
  of	
  meeting	
  another	
  person,	
  sometimes	
  quite	
  unaware	
  of	
  both	
  the	
  
obvious	
   and	
   subtle	
   signs	
   that	
   may	
   be	
   influencing	
   their	
   judgments.	
   Since	
   computers,	
   and	
  
thereby	
  robots	
  in	
  particular,	
  are	
  to	
  some	
  degree	
  treated	
  as	
  social	
  actors,	
  it	
  can	
  be	
  assumed	
  
that	
  people	
  are	
  able	
  to	
  judge	
  robots	
  in	
  a	
  similar	
  way.	
  Jennifer	
  Monahan	
  complemented	
  her	
  
“liking”	
   question	
   with	
   5-­‐point	
   semantic	
   differential	
   scales:	
   nice/awful,	
   friendly/unfriendly,	
  
kind/unkind,	
   and	
   pleasant/unpleasant,	
   because	
   these	
   judgments	
   tend	
   to	
   demonstrate	
  
considerable	
   variance	
   in	
   common	
   with	
   “liking”	
   judgments.	
   Monahan	
   later	
   eliminated	
   the	
  
kind-­‐unkind	
   and	
   pleasant-­‐unpleasant	
   items	
   in	
   her	
   own	
   analysis	
   since	
   they	
   did	
   not	
   load	
  
sufficiently	
   in	
   a	
   factor	
   analysis	
   that	
   also	
   included	
   items	
   from	
   three	
   other	
   factors.	
   The	
  
Cronbach’s	
   Alpha	
   of	
   0.68	
   therefore	
   relates	
   only	
   to	
   this	
   reduced	
   scale.	
   She	
   also	
   included	
  
concepts	
   of	
   physical	
   attraction,	
   conversational	
   skills,	
   and	
   other	
   orientations,	
   which	
   might	
  
become	
   an	
   element	
   of	
   the	
   questionnaire	
   series.	
   In	
   particular,	
   physical	
   attraction	
   might	
  
require	
  additional	
  conceptual	
  and	
  social	
  consideration,	
  since	
  it	
  may	
  also	
  entail	
  sexuality.	
  	
  
	
  
Perceived	
  Safety	
  
Safety	
  is	
  a	
  key	
  issue	
  for	
  robots	
  interacting	
  with	
  humans.	
  The	
  issue	
  has	
  received	
  considerable	
  
attention	
  in	
  the	
  robotics	
  literature,	
  both	
  in	
  systems	
  and	
  standards	
  established	
  for	
  industrial	
  
robots	
  and	
  for	
  service	
  robots	
  intended	
  for	
  use	
  in	
  the	
  home.	
  Examples	
  of	
  design-­‐concerning	
  
category	
  of	
  safety	
  is	
  the	
  mechanical	
  redesign	
  which	
  includes	
  using	
  a	
  whole-­‐body	
  robot	
  visco-­‐
elastic	
  covering,	
  the	
  use	
  of	
  spherical	
  and	
  compliant	
  joints,	
  and	
  distributed	
  parallel	
  actuation	
  
mechanisms	
  to	
  lower	
  the	
  effective	
  inertia	
  of	
  the	
  robot	
  near	
  the	
  end	
  effector.	
  	
  
Perceived	
  safety	
  describes	
  the	
  user’s	
  perception	
  of	
  the	
  level	
  of	
  danger	
  when	
  interacting	
  with	
  
a	
   robot,	
   and	
   the	
   user’s	
   level	
   of	
   comfort	
   during	
   the	
   interaction.	
   Achieving	
   a	
   positive	
  
perception	
  of	
  safety	
  is	
  a	
  key	
  requirement	
  if	
  robots	
  are	
  to	
  be	
  accepted	
  as	
  partners	
  and	
  co-­‐
workers	
   in	
   human	
   environments.	
   Perceived	
   safety	
   and	
   user	
   comfort	
   have	
   rarely	
   been	
  
measured	
  directly.	
  Instead,	
  indirect	
  measures	
  have	
  been	
  used—	
  the	
  measurement	
  of	
  the	
  
affective	
   state	
   of	
   the	
   user	
   through	
   the	
   use	
   of	
   physiological	
   sensors,	
   questionnaires,	
   and	
  
others.	
  That	
  is,	
  instead	
  of	
  asking	
  subjects	
  to	
  evaluate	
  the	
  robot,	
  researchers	
  frequently	
  use	
  
affective	
  state	
  estimation	
  or	
  questionnaires	
  asking	
  how	
  the	
  subject	
  feels	
  in	
  order	
  to	
  measure	
  
the	
  perceived	
  safety	
  and	
  comfort	
  level	
  indirectly.	
  
	
  
Questionnaires	
  can	
  be	
  used	
  to	
  compare	
  different	
  configurations	
  of	
  a	
  robot	
  (Figure	
  11).	
  The	
  
results	
  may	
  then	
  help	
  the	
  developers	
  to	
  choose	
  one	
  option	
  over	
  the	
  other.	
  In	
  the	
  future,	
  this	
  
set	
   of	
   questionnaires	
   could	
   be	
   extended	
   to	
   also	
   include	
   the	
   believability	
   of	
   a	
   robot,	
   the	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.20	
  /	
  56	
  
	
  
enjoyment	
  of	
  interacting	
  with	
  it,	
  and	
  the	
  robot’s	
  social	
  presence.	
  However,	
  the	
  perceptions	
  
of	
  humans	
  are	
  not	
  stable.	
  The	
  more	
  humans	
  get	
  used	
  to	
  the	
  presence	
  of	
  robots,	
  the	
  more	
  
their	
  knowledge	
  and	
  expectations	
  might	
  change.	
  The	
  questionnaires	
  can	
  therefore	
  only	
  offer	
  
a	
  snapshot	
  and	
  it	
  is	
  likely	
  that	
  if	
  the	
  experiment	
  would	
  be	
  repeated	
  in	
  twenty	
  years,	
  it	
  would	
  
yield	
  different	
  results.	
  [17]	
  
	
  
	
  
Figure	
  11.	
  An	
  example	
  of	
  a	
  measurement	
  questionnaire	
  about	
  robot’s	
  factors	
  
	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.21	
  /	
  56	
  
	
  
Below	
  the	
  results	
  of	
  a	
  questionnaire	
  on	
  robot	
  acceptance	
  are	
  extracted.	
  
	
  
An	
   interesting	
   observation	
   they	
   noticed	
   is	
   a	
   tendency	
   towards	
   considering	
   a	
   display	
   of	
   a	
  
facial	
  expression	
  as	
  relatively	
  unimportant.	
  More	
  physical	
  factors	
  such	
  as	
  size	
  and	
  degree	
  of	
  
anthropomorphic	
   design	
   seem	
   to	
   be	
   regarded	
   as	
   neither	
   extremely	
   important	
   nor	
  
unimportant.	
  
From	
   another	
   part	
   of	
   the	
   questionnaire	
   the	
   following	
   properties	
   seem	
   to	
   have	
   some	
  
inherent	
  importance:	
  
• The	
  robot	
  should	
  be	
  a	
  multipurpose	
  tool;	
  
• Easy	
  to	
  instruct;	
  
• It	
  has	
  to	
  behave	
  correctly;	
  
• It	
  should	
  be	
  safe	
  to	
  use	
  and	
  induce	
  confidence;	
  
• Properties	
  seem	
  to	
  be	
  of	
  lesser	
  importance:	
  
• Having	
  a	
  personality;	
  
• Being	
  humanlike;	
  
• Engage	
  in	
  social	
  contact;	
  
	
  
Among	
   the	
   more	
   indecisive	
   properties	
   as:	
   "Autonomy",	
   "Intelligence",	
   and	
   "Appropriate	
  
Size"	
  there	
  are	
  no	
  indications	
  of	
  a	
  general	
  trend.	
  
	
  
It	
  is	
  important	
  to	
  note	
  that	
  a	
  low	
  score	
  does	
  not	
  disqualify	
  a	
  certain	
  property	
  as	
  unimportant,	
  
but	
  rather	
  that	
  the	
  ones	
  with	
  a	
  general	
  higher	
  score	
  are	
  judged	
  as	
  being	
  more	
  important	
  for	
  
instigating	
  a	
  positive	
  feeling	
  towards	
  the	
  robot.	
  In	
  the	
  next	
  part	
  of	
  the	
  questionnaire	
  they	
  
turned	
  the	
  issue	
  around,	
  asking	
  for	
  the	
  properties	
  that	
  were	
  most	
  important	
  in	
  creating	
  a	
  
negative	
  feeling	
  towards	
  the	
  robot.	
  The	
  properties	
  chosen	
  were	
  in	
  this	
  part	
  also	
  chosen	
  for	
  
their	
  possible	
  negative	
  connotation.	
  
	
  
The	
  observable	
  tendencies	
  are	
  in	
  this	
  case	
  fewer	
  and	
  less	
  obvious.	
  The	
  clearest	
  triggers	
  of	
  
negative	
  feelings	
  resulted	
  to	
  the	
  following	
  properties:	
  
• It	
  does	
  not	
  understand	
  the	
  user;	
  
• It	
  is	
  difficult	
  to	
  understand	
  its	
  actions;	
  
• It	
  is	
  complicated	
  to	
  use;	
  
Less	
  clear,	
  but	
  still	
  given	
  high	
  grades	
  by	
  several	
  subjects	
  were	
  the	
  following	
  factors:	
  
• The	
  robot	
  stops	
  unexpectedly;	
  
• The	
  robot	
  looks	
  heavy;	
  
• The	
  robot	
  often	
  asks	
  the	
  user	
  how	
  to	
  proceed;	
  
	
  
More	
  astonishing	
  might	
  be	
  that	
  a	
  machine-­‐like	
  appearance	
  of	
  the	
  robot	
  seems	
  to	
  generate	
  
more	
  positive	
  feelings	
  than	
  a	
  human-­‐like	
  or	
  an	
  animal-­‐like.	
  
Overall	
  most	
  of	
  the	
  factors	
  in	
  this	
  part	
  of	
  the	
  questionnaire	
  were	
  indicated	
  to	
  be	
  important	
  
factors,	
   and	
   there	
   was	
   no	
   real	
   visible	
   trend	
   to	
   judge	
   any	
   one	
   property	
   as	
   more	
   or	
   less	
  
important.	
  The	
  interpretation	
  of	
  this	
  is	
  that	
  most	
  of	
  these	
  properties	
  will	
  give	
  the	
  user	
  a	
  
negative	
  feeling	
  towards	
  the	
  robot,	
  unless	
  they	
  are	
  tended	
  to	
  in	
  the	
  design.	
  [18]	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.22	
  /	
  56	
  
	
  
Research	
  conclusions	
  
	
  
	
  
Human-­‐robot	
  interaction	
  (HRI)	
  is	
  a	
  fairly	
  new	
  branch	
  of	
  HCI	
  (human-­‐computer	
  interaction)	
  
and	
   has	
   gained	
   a	
   lot	
   of	
   attention	
   recently.	
   Concerning	
   a	
   mobile	
   service	
   robot,	
   additional	
  
aspects	
  with	
  respect	
  to	
  user	
  acceptance	
  and	
  their	
  expectations	
  have	
  to	
  be	
  considered.	
  So,	
  
what	
  are	
  people’s	
  views	
  on	
  the	
  role	
  of	
  an	
  intelligent	
  service	
  robot	
  in	
  their	
  home?	
  	
  Different	
  
studies	
   have	
   been	
   conducted	
   to	
   investigate	
   people’s	
   attitudes	
   towards	
   domestic	
   robots.	
  
Syrdal	
  [19]	
  carried	
  out	
  a	
  survey	
  in	
  order	
  to	
  examine	
  adults’	
  attitudes	
  towards	
  an	
  intelligent	
  
service	
  robot.	
  Participants	
  were	
  21-­‐60	
  years	
  old,	
  while	
  most	
  of	
  them	
  were	
  in	
  the	
  age	
  of	
  21-­‐
30.	
   Results	
   show	
   that	
   most	
   of	
   the	
   participants	
   were	
   positive	
   towards	
   the	
   idea	
   of	
   an	
  
intelligent	
  service	
  robot	
  and	
  view	
  it	
  as	
  a	
  domestic	
  machine	
  or	
  smart	
  intelligent	
  equipment	
  
that	
  can	
  be	
  ‘controlled’,	
  but	
  is	
  intelligent	
  enough	
  to	
  perform	
  typical	
  household	
  tasks.	
  On	
  the	
  
other	
  hand,	
  Scopelliti	
  [20]	
  investigated	
  people’s	
  representation	
  of	
  domestic	
  robots	
  across	
  
three	
  different	
  generations	
  and	
  found	
  that	
  while	
  young	
  people	
  tend	
  to	
  have	
  positive	
  feelings	
  
towards	
  domestic	
  robots,	
  elderly	
  people	
  were	
  more	
  frightened	
  of	
  the	
  prospect	
  of	
  a	
  robot	
  in	
  
the	
   home.	
   Studies	
   within	
   the	
   European	
   project	
   COGNIRON	
   assessed	
   people’s	
   attitudes	
  
towards	
   robots	
   via	
   questionnaires	
   following	
   live	
   human-­‐robot	
   interaction	
   trials	
   [21].	
  
Responses	
   from	
   28	
   adults	
   (the	
   majority	
   in	
   the	
   age	
   range	
   26-­‐45)	
   indicated	
   that	
   a	
   large	
  
proportion	
  of	
  participants	
  were	
  in	
  favor	
  of	
  a	
  robot	
  companion,	
  but	
  would	
  prefer	
  it	
  to	
  have	
  a	
  
role	
  of	
  an	
  assistant	
  (79%),	
  machine/appliance	
  (71%)	
  or	
  servant	
  (46%).	
  Few	
  wanted	
  a	
  robot	
  
companion	
  to	
  be	
  a	
  ‘friend’.	
  The	
  majority	
  of	
  the	
  participants	
  wanted	
  the	
  robot	
  to	
  be	
  able	
  to	
  
do	
   household	
   tasks.	
   Also,	
   participants	
   preferred	
   a	
   robot	
   that	
   is	
   predictable,	
   controllable,	
  
considerate	
   and	
   polite.	
   Humanlike	
   communication	
   was	
   desired	
   for	
   a	
   robot	
   companion,	
  
however,	
   human-­‐like	
   behavior	
   and	
   appearance	
   were	
   less	
   important.	
   These	
   three	
   studies,	
  
conducted	
   in	
   different	
   European	
   countries,	
   agreed	
   with	
   respect	
   to	
   the	
   desired	
   role	
   of	
   a	
  
service	
  robot	
  in	
  the	
  home:	
  an	
  assistant	
  able	
  to	
  carry	
  out	
  useful	
  tasks,	
  and	
  not	
  necessarily	
  a	
  
‘friend’	
  with	
  human-­‐like	
  appearance.	
  These	
  considerations	
  led	
  to	
  the	
  definition	
  of	
  a	
  robot	
  
companion	
  which	
  must	
  a)	
  be	
  able	
  to	
  perform	
  a	
  range	
  of	
  useful	
  tasks	
  or	
  functions,	
  and	
  b)	
  
carry	
  out	
  these	
  tasks	
  or	
  functions	
  in	
  a	
  manner	
  that	
  is	
  socially	
  acceptable	
  and	
  comfortable	
  for	
  
people	
   it	
   shares	
   the	
   environment	
   with	
   and/or	
   it	
   interacts	
   with.	
   [19]	
   This	
   creates	
   the	
  
following	
   challenge	
   for	
   the	
   development	
   of	
   such	
   a	
   robot:	
   we	
   have	
   to	
   bridge	
   the	
   gap	
  
between	
   functionality,	
   which	
   goes	
   along	
   with	
   hard	
   technological	
   properties	
   of	
   e.g.	
   an	
  
industrial	
  robot,	
  and	
  social	
  acceptance,	
  which	
  goes	
  along	
  with	
  the	
  comfortable	
  design	
  of	
  
e.g.	
  an	
  electronic	
  pet.	
  [16]	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.23	
  /	
  56	
  
	
  
The	
  aim	
  
	
  
	
  
The	
  aim	
  of	
  this	
  thesis	
  is	
  to	
  investigate	
  the	
  user’s	
  familiarity	
  of	
  different	
  design	
  alternatives	
  
for	
  service	
  robots	
  according	
  to	
  the	
  “uncanny	
  valley”	
  hypothesis.	
  
For	
   this	
   purpose	
   a	
   number	
   of	
   approximately	
   seven	
   factor	
   levels	
   of	
   the	
   robot	
   appearance	
  
varying	
   in	
   human-­‐likeness	
   were	
   designed.	
   Also	
   a	
   questionnaire	
   comparing	
   the	
   different	
  
alternatives	
  was	
  created	
  for	
  the	
  test	
  subjects.	
  An	
  empirical	
  study	
  design	
  was	
  developed	
  and	
  
a	
  pretest	
  was	
  carried	
  out.	
  Pretest	
  results	
  of	
  the	
  influences	
  of	
  the	
  familiarity	
  and	
  likeness	
  will	
  
be	
   given	
   in	
   a	
   graph.	
   Then	
   a	
   comparison	
   between	
   the	
   resulted	
   graph	
   and	
   the	
   one	
   of	
   the	
  
uncanny	
  valley	
  completes	
  the	
  thesis.	
  
	
  
Design	
  concept	
  
	
  
	
  
The	
   concept	
   of	
   the	
   design	
   approach	
   in	
   this	
   study	
   is	
   based	
   on	
   the	
   idea	
   to	
   cover	
   as	
   many	
  
significant	
   points	
   of	
   the	
   uncanny	
   valley	
   graph	
   as	
   possible	
   (Figure	
   12.).	
   For	
   each	
   point	
   a	
  
different	
  appearance	
  alternative	
  is	
  designed.	
  Starting	
  with	
  an	
  industrial	
  machine-­‐like	
  robot,	
  
each	
  gets	
  more	
  human-­‐like	
  than	
  the	
  previous.	
  This	
  is	
  the	
  expected	
  ranking	
  order	
  given	
  to	
  
them	
  by	
  the	
  authors,	
  according	
  to	
  the	
  “uncanny	
  valley”	
  hypothesis.	
  	
  	
  	
  	
  
Research	
  questions	
  
Related	
   to	
   the	
   above	
   issues,	
   the	
   present	
   study	
   addressed	
   the	
   following	
   main	
   research	
  
questions:	
  	
  
What	
  is	
  the	
  importance	
  of	
  robot	
  appearance	
  for	
  less	
  human-­‐looking	
  robots?	
  
Do	
  people	
  prefer	
  more	
  human-­‐like	
  appearance	
  in	
  robots	
  that	
  they	
  interact	
  with?	
  
Does	
  gender	
  factor	
  in	
  both	
  sides	
  –	
  robots’	
  and	
  test	
  subjects’,	
  influences	
  the	
  judgments	
  of	
  the	
  
participants?	
  
	
  
How	
  many	
  design	
  alternatives	
  are	
  needed	
  in	
  order	
  to	
  get	
  more	
  complete	
  results?	
  
	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.24	
  /	
  56	
  
	
  
	
  
Figure	
  12.	
  A	
  simplified	
  graph	
  of	
  the	
  uncanny	
  valley,	
  on	
  which	
  the	
  concept	
  of	
  current	
  study	
  is	
  based	
  
	
  
Design	
  requirements	
  and	
  limitations	
  
According	
  to	
  the	
  concept	
  above,	
  the	
  development	
  of	
  robot	
  appearance	
  should	
  follow	
  a	
  line,	
  
instead	
  of	
  a	
  curve,	
  excluding	
  the	
  hypothetic	
  valley	
  of	
  the	
  graph.	
  With	
  this	
  approach	
  in	
  each	
  
design	
  should	
  be	
  added	
  a	
  human-­‐like	
  feature,	
  enough	
  to	
  make	
  it	
  distinguishable	
  as	
  a	
  more	
  
human-­‐like	
  than	
  the	
  previous	
  one	
  (Figure	
  13.).	
  
The	
  graph	
  starts	
  with	
  a	
  robot	
  which	
  is	
  relatively	
  machine-­‐like	
  in	
  appearance.	
  It	
  will	
  have	
  no	
  
overtly	
  human-­‐like	
  features.	
  Then	
  the	
  development	
  will	
  go	
  through	
  a	
  humanoid	
  -­‐	
  a	
  robot	
  
which	
  is	
  not	
  realistically	
  human-­‐like	
  in	
  appearance	
  and	
  is	
  readily	
  perceived	
  as	
  a	
  robot	
  by	
  
human	
  users.	
  However,	
  it	
  may	
  possess	
  some	
  human-­‐like	
  features,	
  which	
  are	
  usually	
  stylized,	
  
simplified	
  or	
  cartoon-­‐like	
  versions	
  of	
  the	
  human	
  equivalents,	
  including	
  some	
  or	
  all	
  of	
  the	
  
following:	
   a	
   head,	
   facial	
   features,	
   eyes,	
   ears,	
   eyebrows,	
   arms,	
   hands,	
   legs.	
   The	
   important	
  
principle	
  that	
  should	
  be	
  kept	
  in	
  the	
  design	
  line	
  is	
  that	
  any	
  feature,	
  if	
  possessed	
  in	
  a	
  robot	
  
appearance,	
  should	
  be	
  more	
  human-­‐like	
  in	
  the	
  next	
  robot.	
  Both	
  may	
  also	
  have	
  wheels	
  for	
  
locomotion	
  or	
  use	
  legs	
  for	
  walking.	
  The	
  graph	
  should	
  end	
  with	
  an	
  android	
  -­‐	
  a	
  robot	
  which	
  
exhibits	
  appearance	
  which	
  is	
  as	
  close	
  to	
  a	
  real	
  human	
  appearance	
  as	
  technically	
  possible.	
  	
  
The	
  working	
  definition	
  of	
  robot	
  appearance	
  for	
  humanoid	
  robots	
  used	
  in	
  this	
  study	
  is	
  based	
  
on	
  the	
  definitions	
  for	
  animated	
  agents’	
  appearances	
  adopted	
  by	
  Gong	
  and	
  Nass	
  [22],	
  and	
  for	
  
android	
  robots	
  from	
  Mac-­‐Dorman	
  and	
  Ishiguro	
  [15].	
  	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.25	
  /	
  56	
  
	
  
	
  
Figure	
  13.	
  Line	
  of	
  design	
  alternatives	
  development	
  
	
  
	
  
For	
   a	
   better	
   user’s	
   recognition	
   each	
   design	
   alternative	
   should	
   be	
   shown	
   in	
   identical	
  
environment	
  performing	
  the	
  same	
  tasks.	
  In	
  this	
  way	
  the	
  robot’s	
  purpose	
  will	
  be	
  clear	
  for	
  the	
  
participants,	
  so	
  they	
  won’t	
  be	
  confused	
  by	
  the	
  different	
  appearances.	
  
The	
   styling	
   of	
   the	
   sketches	
   and	
   the	
   models	
   respectively,	
   incl.	
   colors,	
   shapes,	
   materials,	
  
should	
   be	
   also	
   identical	
   in	
   the	
   developed	
   alternatives,	
   because	
   if	
   not,	
   there	
   might	
   be	
  
unwanted	
  impact	
  on	
  the	
  users’	
  impression	
  about	
  them.	
  For	
  example,	
  if	
  a	
  box	
  shape	
  is	
  used	
  
for	
   a	
   head	
   in	
   the	
   machine-­‐like	
   alternative,	
   then	
   a	
   box	
   shaped	
   head	
   should	
   be	
   used	
   in	
   a	
  
similar	
  way	
  for	
  the	
  more	
  human-­‐like	
  alternative.	
  
	
  
Design	
  project
	
  
Below	
  are	
  the	
  initial	
  sketches	
  for	
  six	
  of	
  the	
  alternatives.	
  The	
  performing	
  task	
  for	
  each	
  one	
  is	
  
fetching	
  and	
  carrying	
  a	
  glass	
  and	
  a	
  bottle	
  of	
  water.	
  	
  
The	
  seventh	
  level	
  of	
  human	
  likeness	
  is	
  actually	
  a	
  real	
  human.	
  Note	
  that	
  for	
  humanoid	
  and	
  
android	
   parts	
   of	
   the	
   graph	
   authors	
   have	
   developed	
   a	
   second	
   level	
   factor	
   alternatives	
   for	
  
each	
   robot	
   –	
   a	
   male	
   and	
   a	
   female	
   gender.	
   	
   This	
   altogether	
   makes	
   eleven	
   alternative	
  
visualizations.	
  	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.26	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  14.	
  Initial	
  sketches	
  for	
  design	
  alternative	
  Nr.1	
  
	
  
Alternative	
  #1	
  
The	
   main	
   concept	
   here	
   is	
   to	
   use	
   basic	
   shapes	
  
(sphere,	
   cylinder)	
   and	
   simple	
   cuts	
   for	
   creating	
   a	
  
robot	
  which	
  is	
  closer	
  to	
  an	
  object	
  executing	
  the	
  task	
  
–	
  carrying	
  a	
  bottle.	
  What	
  is	
  crucial	
  here	
  is	
  that	
  this	
  
object	
  should	
  look	
  like	
  it	
  is	
  moving	
  itself,	
  and	
  not	
  just	
  
standing	
  (as	
  the	
  participants	
  are	
  going	
  to	
  see	
  only	
  a	
  
picture,	
  not	
  an	
  animation).	
  That	
  is	
  why	
  appearance	
  
#1	
   should	
   definitely	
   have	
   wheels,	
   or	
   chains,	
   and	
   if	
  
wheels	
  –	
  then	
  preferably	
  more	
  than	
  two,	
  for	
  more	
  
confidence	
  in	
  its	
  stability.	
  	
  	
  	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.27	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  15.	
  Initial	
  sketches	
  for	
  design	
  alternative	
  Nr.2	
  
Alternative	
  #2	
  
Here	
  the	
  concept	
  resembles	
  the	
  one	
  of	
  Care-­‐o-­‐bot	
  –	
  
the	
   two-­‐side	
   concept	
   (working	
   and	
   serving	
   side).	
  
That	
  means	
  assuming	
  there	
  is	
  at	
  least	
  one	
  hand	
  or	
  
some	
  other	
  part	
  for	
  fetching	
  objects.	
  Still,	
  it	
  is	
  not	
  an	
  
android	
   or	
   a	
   humanoid,	
   but	
   its	
   silhouette	
   could	
  
remind	
  of	
  a	
  short	
  waiter.	
  A	
  display	
  for	
  serving	
  side	
  is	
  
preferable,	
   as	
   well.	
   Movement	
   could	
   be	
   two-­‐	
   or	
  
three-­‐wheels,	
   but	
   again	
   because	
   of	
   stability	
   and	
  
easier	
   recognition,	
   a	
   wheel	
   system	
   in	
   the	
   back	
   can	
  
play	
  the	
  role	
  of	
  stabilizer.	
  	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.28	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  16.	
  Initial	
  sketches	
  for	
  design	
  alternative	
  Nr.3	
  
	
  
Alternative	
  #3	
  
As	
   robots	
   get	
   more	
   and	
   more	
   human-­‐like,	
   in	
  
alternative	
   #3	
   we	
   can	
   already	
   nave	
   some	
   more	
  
human	
   features	
   like	
   head	
   and	
   two	
   arms,	
   for	
  
example.	
   In	
   this	
   case	
   the	
   head	
   is	
   rather	
   a	
   helmet,	
  
and	
  a	
  big	
  mask	
  instead	
  of	
  a	
  display	
  or	
  face.	
  Two	
  arms	
  
or	
  a	
  tray,	
  attached	
  to	
  the	
  body	
  is	
  the	
  functional	
  part	
  
which	
   the	
   patient	
   interacts	
   with.	
   Proportions	
   are	
  
childish	
  –	
  bigger	
  head	
  and	
  smaller	
  body.	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.29	
  /	
  56	
  
	
  
	
  
Figure	
  17.	
  Initial	
  sketches	
  for	
  design	
  alternative	
  Nr.4	
  
	
  
Alternative	
  #4	
  
The	
  concept	
  is	
  developing	
  smoothly	
  to	
  a	
  humanoid,	
  
adding	
   legs,	
   which	
   may	
   be	
   separate,	
   or	
   just	
  
imitating.	
  Here	
  we	
  can	
  already	
  have	
  a	
  neck,	
  attached	
  
to	
   a	
   head	
   or	
   a	
   helmet.	
   Smaller	
   mask	
   and	
   more	
  
detailed	
  arms,	
  as	
  well	
  as	
  shoes/feet	
  are	
  preferable.	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.30	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  18.	
  Initial	
  sketches	
  for	
  design	
  alternative	
  Nr.4,	
  part	
  2	
  
	
  
Alternative	
  #4	
  
The	
   concept	
   of	
   alternatives	
   from	
   #4	
   onwards	
  
will	
  differ	
  in	
  genders,	
  as	
  well.	
  In	
  this	
  design	
  the	
  
concept	
   changed	
   its	
   direction	
   to	
   an	
   animation	
  
character’s	
  proportion	
  for	
  the	
  humanoids.	
  This	
  
means	
   really	
   exaggerated	
   big	
   head	
   and	
  
significantly	
   smaller	
   body	
   in	
   different	
   from	
  
human	
  proportions.	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.31	
  /	
  56	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
Figure	
  19.	
  Initial	
  sketches	
  for	
  design	
  alternative	
  Nr.5	
  
	
  
Alternative	
  #5	
  
Approaching	
  the	
  android	
  we	
  risk	
  falling	
  into	
  the	
  uncanny	
  valley.	
  That’s	
  why	
  we	
  want	
  to	
  make	
  
this	
  falling	
  controlled,	
  by	
  switching	
  to	
  exaggerated,	
  but	
  rather	
  detailed,	
  human	
  features,	
  but	
  
still	
   following	
   the	
   line	
   of	
   human	
   likeness.	
   For	
   example	
   –	
   making	
   the	
   eyes	
   and	
   the	
   mouth	
  
extremely	
  big	
  is	
  acceptable,	
  but	
  placing	
  the	
  face	
  on	
  the	
  body	
  part	
  is	
  not	
  –	
  it	
  is	
  less	
  human	
  like	
  
than	
  the	
  previous	
  alternatives.	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.32	
  /	
  56	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
	
  
Figure	
  20.	
  Initial	
  sketches	
  for	
  design	
  alternative	
  Nr.6	
  
	
   	
  
Alternative	
  #6	
  
For	
  the	
  right	
  hill	
  of	
  the	
  uncanny	
  valley	
  graph	
  we	
  have	
  
and	
   android	
   which	
   should	
   possess	
   already	
   human-­‐
like	
   detailed	
   features	
   like	
   real-­‐size	
   eyes	
   and	
   an	
  
appropriate	
  nice	
  sight.	
  Nose	
  and	
  ears,	
  as	
  well	
  as	
  an	
  
open	
  mouth	
  can	
  finish	
  the	
  face.	
  Their	
  arms	
  and	
  feet	
  
can	
  be	
  supplied	
  with	
  fingers,	
  instead	
  of	
  just	
  clippers.	
  
Longer	
  legs	
  and	
  knees	
  would	
  increase	
  the	
  impression	
  
of	
  available	
  realistic	
  movement.	
  	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.33	
  /	
  56	
  
	
  
Pretest	
  design
	
  
Scenario	
  
The	
   context	
   chosen	
   for	
   the	
   study	
   and	
   associated	
   HRI	
   trials	
   was	
   that	
   of	
   a	
   domestic	
   robot	
  
bringing	
  a	
  glass	
  and	
  a	
  bottle	
  of	
  water	
  to	
  the	
  participant.	
  	
  
Participants	
  and	
  procedure	
  
At	
  the	
  beginning	
  of	
  each	
  trial,	
  information	
  about	
  the	
  purpose	
  of	
  the	
  current	
  test	
  was	
  given	
  
to	
  each	
  of	
  the	
  participants	
  and	
  detailed	
  instructions	
  for	
  participating	
  in	
  the	
  experiment.	
  A	
  
supervisor	
  was	
  on	
  hand	
  to	
  answer	
  any	
  further	
  questions	
  and	
  to	
  repeat	
  the	
  instructions	
  if	
  
necessary.	
  	
  
Twenty	
  participants	
  took	
  part	
  in	
  the	
  pretests.	
  There	
  were	
  mainly	
  students	
  (60%),	
  7	
  female	
  
and	
   13	
   male.	
   The	
   average	
   age	
   was	
   26.	
   Everyone	
   could	
   communicate	
   in	
   English	
   with	
   12	
  
growing	
  up	
  in	
  Germany,	
  and	
  8	
  growing	
  up	
  in	
  other	
  non-­‐English	
  speaking	
  countries	
  (4	
  in	
  India,	
  
3	
  in	
  Spain,	
  1	
  in	
  Turkey).	
  The	
  majority	
  (80%)	
  said	
  they	
  had	
  some	
  information	
  about	
  any	
  kind	
  
of	
  robots	
  –	
  incl.	
  rescue	
  robots,	
  service	
  robots,	
  industrial	
  building	
  robots,	
  but	
  only	
  30%	
  of	
  
them	
  had	
  real	
  live	
  experience	
  with	
  robots.	
  The	
  participants	
  were	
  all	
  volunteers	
  and	
  none	
  
received	
  remuneration.	
  	
  
The	
  pretests	
  took	
  place	
  in	
  a	
  university	
  laboratory.	
  The	
  test	
  subjects	
  were	
  sitting	
  in	
  front	
  of	
  
24”	
  LCD	
  screen.	
  Only	
  one	
  subject	
  at	
  a	
  time	
  was	
  examined.	
  Each	
  of	
  them	
  completed	
  the	
  first	
  
part	
   of	
   a	
   questionnaire	
   individually,	
   providing	
   basic	
   demographic	
   details	
   including	
  
background,	
  gender,	
  and	
  age,	
  as	
  well	
  as	
  robot	
  information	
  or	
  interaction	
  experience	
  before	
  
they	
   were	
   exposed	
   to	
   the	
   testing.	
   The	
   relevant	
   questions	
   from	
   the	
   questionnaire	
   are	
  
provided	
   below	
   in	
   the	
   Questionnaire	
   section.	
   The	
   participants	
   were	
   then	
   shown	
  
visualizations	
  of	
  eleven	
  robots.	
  Robots’	
  appearances	
  were	
  labeled	
  with	
  numbers	
  only	
  for	
  the	
  
supervisor	
   in	
   order	
   to	
   avoid	
   any	
   possible	
   influences	
   on	
   the	
   subjects’	
   judgments.	
   Each	
  
participant	
  was	
  given	
  a	
  different	
  sequence	
  of	
  the	
  robots’	
  pictures.	
  In	
  this	
  way	
  there	
  is	
  no	
  
possibility	
  for	
  them	
  to	
  guess	
  our	
  predicted	
  initial	
  ranking	
  order	
  of	
  the	
  alternatives	
  as	
  they	
  
were	
   created.	
   The	
   participants	
   had	
   to	
   fill	
   in	
   the	
   second	
   part	
   of	
   the	
   questionnaire	
   while	
  
watching	
  the	
  pictures	
  in	
  order	
  to	
  mark	
  their	
  preferences	
  towards	
  each	
  robot	
  appearance.	
  A	
  
sample	
  of	
  the	
  test	
  sheet	
  is	
  shown	
  in	
  the	
  Questionnaire	
  section,	
  as	
  well.	
  Each	
  of	
  the	
  two	
  
20cm-­‐long	
  horizontal	
  lines	
  represents	
  a	
  100%	
  scale	
  for	
  both	
  factors	
  –	
  Human	
  likeness	
  and	
  
Familiarity.	
  For	
  the	
  upper	
  one	
  (Human	
  likeness)	
  the	
  extremities	
  are:	
  0%	
  for	
  Machine-­‐like,	
  
and	
   100%	
   for	
   Human	
   like	
   appearance	
   of	
   the	
   visualized	
   robot.	
   For	
   the	
   second	
   line,	
  
respectively:	
  0%	
  for	
  Very	
  strange,	
  and	
  100%	
  for	
  Very	
  familiar.	
  Each	
  participant	
  had	
  to	
  mark	
  
(using	
  a	
  vertical	
  line	
  or	
  a	
  cross)	
  on	
  each	
  line	
  their	
  evaluation.	
  	
  
Finally,	
   the	
   participants	
   were	
   asked	
   whether	
   they	
   had	
   any	
   difficulty	
   completing	
   the	
  
questionnaire	
  or	
  understanding	
  what	
  they	
  were	
  supposed	
  to	
  do	
  in	
  the	
  experiment;	
  and	
  what	
  
their	
  impression	
  was	
  of	
  the	
  11	
  robot	
  designs.	
  	
  
All	
   evaluations	
   from	
   the	
   second	
   part	
   of	
   the	
   questionnaire	
   were	
   then	
   measured	
   in	
  
centimeters	
  and	
  the	
  data	
  was	
  transferred	
  into	
  a	
  table	
  (Table	
  1.).	
  	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.34	
  /	
  56	
  
	
  
Visualizations	
  
Visualizations	
  were	
  designed	
  and	
  rendered	
  in	
  CAD	
  software	
  –	
  3D	
  Studio	
  Max	
  Design	
  2012.	
  
At	
  first,	
  all	
  the	
  robots	
  should	
  be	
  visualized	
  in	
  identical	
  environment	
  and	
  doing	
  the	
  same	
  task	
  
–	
  in	
  this	
  case	
  they	
  carry	
  a	
  tray	
  with	
  a	
  bottle	
  and	
  a	
  glass	
  of	
  water.	
  According	
  to	
  some	
  findings	
  
of	
   Walters	
   et	
   al.	
   [23],	
   humans	
   strongly	
   did	
   not	
   like	
   a	
   direct	
   frontal	
   approach	
   by	
   a	
   robot,	
  
especially	
   while	
   sitting	
   (even	
   at	
   a	
   table)	
   or	
   while	
   standing	
   with	
   their	
   back	
   to	
   a	
   wall.	
   An	
  
approach	
  from	
  the	
  front	
  left	
  or	
  front	
  right	
  was	
  preferred.	
  In	
  this	
  case,	
  all	
  the	
  visualizations	
  
are	
  taken	
  with	
  the	
  robots	
  “approaching”	
  from	
  the	
  front	
  right	
  side	
  of	
  the	
  screen.	
  	
  	
  
Design	
  alternative	
  #1	
  (Figure	
  21.)	
  has	
  been	
  chosen	
  to	
  resemble	
  a	
  simplistic	
  association	
  with	
  
the	
  famous	
  R2D2	
  robot.	
  This	
  is	
  because	
  it	
  is	
  familiar	
  enough,	
  and	
  still	
  does	
  not	
  possess	
  any	
  
human	
  features,	
  but	
  rather	
  is	
  closer	
  to	
  an	
  object	
  from	
  a	
  living	
  environment	
  (e.g.	
  a	
  small	
  table	
  
or	
  a	
  mobile	
  mini-­‐bar).	
  	
  The	
  function	
  concept	
  behind	
  is	
  that	
  the	
  hemisphere	
  on	
  the	
  top	
  can	
  
open	
  from	
  the	
  needed	
  side	
  (either	
  from	
  user’s	
  side,	
  or	
  working	
  side	
  –	
  where	
  bottle	
  charging	
  
happens).	
  The	
  robot	
  moves	
  by	
  3	
  spherical	
  wheels,	
  symmetrically	
  placed	
  on	
  its	
  bottom.	
  
Design	
  alternative	
  #2	
  (Figure	
  22.)	
  won	
  among	
  the	
  rest	
  of	
  the	
  sketch	
  concepts	
  because	
  its	
  lack	
  
of	
  a	
  hand	
  makes	
  the	
  curve	
  of	
  transition	
  between	
  alternatives	
  smoother.	
  Still	
  its	
  silhouette	
  
and	
   movement	
   signs	
   remind	
   more	
   of	
   an	
   object	
   than	
   a	
   human.	
   It	
   possesses	
   a	
   display	
   for	
  
interaction	
   with	
   the	
   user.	
   The	
   wheels	
   are	
   also	
   three,	
   but	
   cylindrical	
   and	
   asymmetrically	
  
places,	
  two	
  of	
  which	
  in	
  the	
  back.	
  This	
  was	
  imposed	
  by	
  the	
  asymmetrical	
  shape,	
  which	
  makes	
  
the	
  silhouette	
  to	
  have	
  a	
  direction	
  facing	
  the	
  user	
  or	
  the	
  working	
  plot.	
  In	
  non-­‐active	
  state	
  the	
  
display	
  can	
  be	
  closed.	
  
Design	
   alternative	
   #3	
   (Figure	
   23.)	
   is	
   designed	
   to	
   have	
   rather	
   baby	
   proportions	
   –	
   very	
   big	
  
head	
   and	
   a	
   small	
   body,	
   short	
   arms	
   and	
   posing	
   like	
   squatting	
   or	
   crawling.	
   That’s	
   why	
   the	
  
wheels	
  here	
  are	
  already	
  four	
  (association	
  with	
  a	
  baby	
  crawling).	
  Its	
  spherical	
  helmet	
  play	
  the	
  
role	
  of	
  a	
  head,	
  together	
  with	
  a	
  big	
  mask,	
  looks	
  more	
  like	
  a	
  space	
  suit.	
  The	
  tray	
  in	
  this	
  case	
  is	
  
attached	
  to	
  two	
  short	
  “arms”	
  which	
  could	
  be	
  height-­‐adjusted.	
  
Design	
   alternatives	
   #4	
   (Figures	
   24.25.)	
   are	
   already	
   in	
   two	
   versions	
   –	
   female	
   and	
   male.	
  
Cartoons	
  –	
  a	
  bigger	
  head	
  and	
  tiny	
  body,	
  inspire	
  their	
  proportions.	
  The	
  head	
  is	
  still	
  spherical	
  
in	
  shape,	
  keeping	
  the	
  style	
  from	
  the	
  previous	
  alternative	
  #3.	
  There	
  is	
  a	
  darker	
  helmet,	
  which	
  
can	
  be	
  considered	
  as	
  hair.	
  Here	
  the	
  mask	
  is	
  smaller,	
  resembling	
  sunglasses,	
  slightly	
  different	
  
depending	
  on	
  the	
  gender.	
  Mouth	
  appears	
  on	
  the	
  face	
  to	
  make	
  it	
  more	
  character	
  look-­‐alike.	
  
More	
  humanoid	
  features	
  come	
  up	
  –	
  arms	
  are	
  proportional	
  to	
  the	
  body	
  and	
  are	
  separate	
  
from	
   the	
   tray.	
   The	
   hands	
   are	
   like	
   clippers.	
   The	
   female	
   has	
   a	
   “dress”,	
   while	
   the	
   male	
   is	
  
“dressed”	
  in	
  trousers.	
  Their	
  moving	
  equipment	
  consists	
  of	
  small	
  “shoes”,	
  and	
  it	
  is	
  supposed	
  
to	
  slide	
  on	
  the	
  ground,	
  instead	
  of	
  rolling	
  wheels.	
  
Design	
   alternatives	
   #5	
   (Figures	
   26.27.)	
   go	
   closer	
   to	
   the	
   predicted	
   uncanny	
   valley	
   of	
   the	
  
graph.	
   That’s	
   why	
   the	
   concept	
   of	
   exaggerating	
   human	
   features	
   is	
   applied	
   here.	
   It	
   is	
  
expressed	
   most	
   noticeable	
   in	
   face	
   features	
   like	
   bigger	
   eyes	
   and	
   wider	
   mouth,	
   small	
   and	
  
animal-­‐type	
  nose.	
  Gender	
  signs	
  are	
  obvious	
  on	
  the	
  upper	
  bodies.	
  Limbs	
  are	
  turned	
  up	
  side	
  
down	
   –	
   wider	
   in	
   the	
   lower	
   parts.	
   But	
   the	
   extremities	
   here	
   are	
   more	
   detailed	
   than	
   the	
  
previous	
  alternative	
  –	
  they	
  have	
  knees	
  and	
  elbows,	
  as	
  they	
  are	
  even	
  moving	
  independently	
  
(not	
  sliding	
  together).	
  	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.35	
  /	
  56	
  
	
  
Design	
  alternatives	
  #6	
  (Figures	
  28.29.)	
  are	
  going	
  up	
  to	
  the	
  android	
  on	
  the	
  predicted	
  graph	
  in	
  
this	
  study.	
  That’s	
  why	
  face	
  features	
  are	
  much	
  closer	
  to	
  the	
  real	
  human	
  eyes,	
  mouth,	
  nose,	
  
and	
   even	
   cheeks,	
   as	
   well	
   as	
   the	
   head	
   shape,	
   which	
   is	
   already	
   looking	
   like	
   human	
   skull.	
  
Eyebrows	
  and	
  ears	
  are	
  added.	
  The	
  extremities	
  are	
  in	
  proper	
  human	
  proportions.	
  Hands	
  have	
  
fingers	
  and	
  feet	
  are	
  more	
  human-­‐like.	
  	
  
Design	
  alternatives	
  #7	
  (Figures	
  30.31.)	
  are	
  supposed	
  to	
  be	
  the	
  real	
  humans.	
  They	
  are	
  re-­‐
defined	
   three-­‐dimensional	
   models,	
   adapted	
   to	
   the	
   style	
   of	
   the	
   previous	
   alternatives.	
   The	
  
most	
   important	
   here	
   was	
   their	
   sight	
   –	
   it	
   should	
   be	
   nice	
   and	
   friendly,	
   not	
   frightening	
   the	
  
users.	
  This	
  task	
  is	
  very	
  difficult,	
  because	
  it	
  is	
  still	
  not	
  a	
  real	
  human,	
  and	
  it	
  is	
  obvious	
  for	
  the	
  
test	
  subject,	
  who	
  is	
  a	
  real	
  human.	
  Both	
  female	
  and	
  male	
  models	
  present	
  nice	
  looking	
  waiters	
  
or	
  companion	
  staff.	
  Their	
  clothes	
  are	
  in	
  same	
  colors	
  like	
  the	
  robots,	
  so	
  test	
  subject	
  would	
  
not	
  be	
  confused	
  of	
  too	
  different	
  factor	
  levels	
  (like	
  different	
  colors).	
  	
  	
  
	
  	
  	
  
	
  
	
  
Figure	
  21.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.1	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.36	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  22.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.2	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.37	
  /	
  56	
  
	
  
	
  
Figure	
  23.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.3	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.38	
  /	
  56	
  
	
  
	
  
Figure	
  24.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.4	
  –	
  female	
  version	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.39	
  /	
  56	
  
	
  
	
  
Figure	
  25.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.4	
  –	
  male	
  version	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.40	
  /	
  56	
  
	
  
	
  
Figure	
  26.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.5	
  –	
  female	
  version	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.41	
  /	
  56	
  
	
  
	
  
Figure	
  27.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.5	
  –	
  male	
  version	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.42	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  28.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.6	
  –	
  female	
  version	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.43	
  /	
  56	
  
	
  
	
  
	
  
Figure	
  29.	
  3-­‐dimensional	
  model	
  for	
  design	
  alternative	
  Nr.6	
  –male	
  version	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.44	
  /	
  56	
  
	
  
	
  
Figure	
  30.	
  3-­‐dimensional	
  model	
  for	
  the	
  real	
  human	
  –	
  female	
  version	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.45	
  /	
  56	
  
	
  
	
  
Figure	
  31.	
  3-­‐dimensional	
  model	
  for	
  the	
  real	
  human	
  –male	
  version	
  
	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.46	
  /	
  56	
  
	
  
Questionnaire	
  
	
  
	
  
no.	
  of	
  participant:	
  	
  	
  	
   	
   	
   date:	
  	
  	
  	
   	
   	
   time:	
  
	
  
First,	
  we	
  ask	
  you	
  for	
  some	
  information	
  about	
  yourself.	
  Thank	
  you!	
  
	
  
1.	
  How	
  old	
  are	
  you?	
  
______	
  years	
  old	
  
	
  
2.	
  What	
  is	
  your	
  gender?	
  
□	
  Male	
  	
  	
  	
  □	
  Female	
  
	
  
3.	
  What	
  is	
  your	
  highest	
  level	
  of	
  education?	
  
□	
  No	
  graduation	
  □	
  secondary	
  modern	
  school	
  	
  	
  	
  □	
  PhD	
  	
  
□	
  Business	
  school	
  □	
  On-­‐the-­‐job	
  training	
  	
  	
  	
  	
  	
  □	
  A-­‐level	
  
□	
  Studies	
  	
  	
  	
  	
  □	
  secondary	
  school	
  level	
  I	
  certificate	
  □	
  Other:	
  _________	
  
	
  
4.	
  Where	
  do	
  you	
  work	
  /	
  you	
  have	
  worked?	
  
□	
  Craft	
  	
  	
  	
  	
  	
  	
  	
  □	
  Social-­‐/Humanities	
  sciences	
  
□	
  Mercantile	
  /Administration	
  □	
  technical/Natural	
  science	
  
□	
  Housewife/-­‐husband	
  	
  	
  	
  □	
  Other:	
  _______________________	
  
	
  
5.	
  Have	
  you	
  ever	
  read,	
  heard	
  or	
  seen	
  anything	
  about	
  robots	
  	
  
(e.g.	
  in	
  newspapers,	
  	
  books,	
  television,	
  movies	
  or	
  from	
  friends)?	
  
□	
  No	
  
□	
  Yes,	
  about	
  following	
  robots	
  	
  __________________________________	
  
__________________________________	
  
__________________________________	
  
__________________________________	
  
	
  
6.	
  Did	
  you	
  have	
  already	
  gained	
  some	
  experience	
  by	
  using	
  robots?	
  
□	
  No	
  	
  	
  □	
  Yes,	
  with	
  	
  	
  	
  __________________________________	
  
__________________________________	
  
__________________________________	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.47	
  /	
  56	
  
	
  
	
  
	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.48	
  /	
  56	
  
	
  
Pretest	
  results	
  and	
  analyses
	
  
Below	
  all	
  the	
  data	
  collected	
  from	
  the	
  pretests	
  is	
  extracted	
  (Table	
  1.).	
  	
  
First,	
  the	
  average	
  ratings	
  for	
  Human	
  likeness	
  and	
  Familiarity	
  were	
  measured	
  (Figure	
  32.).	
  
The	
   standard	
   deviations	
   are	
   also	
   shown	
   there.	
   In	
   our	
   case	
   there	
   are	
   some	
   big	
   values	
   for	
  
some	
  of	
  the	
  alternatives	
  (mostly	
  for	
  extremities	
  –	
  the	
  machine	
  like	
  and	
  the	
  real	
  humans).	
  
Regarding	
  the	
  low	
  numbers	
  of	
  test	
  subjects	
  in	
  the	
  pretests,	
  we’d	
  rather	
  conclude	
  that	
  the	
  
standard	
   deviation	
   shows	
   us	
   that	
   people	
   have	
   very	
   different	
   opinions	
   about	
   the	
   robots’	
  
appearance.	
  These	
  deviations	
  could	
  be	
  in	
  a	
  different	
  range	
  after	
  the	
  real	
  tests	
  later.	
  
	
  
Figure	
  32.	
  Human	
  likeness	
  and	
  Familiarity	
  curves	
  with	
  Standard	
  deviations	
  measured	
  on	
  11-­‐point	
  scale	
  
	
  
	
  
An	
  important	
  data	
  about	
  the	
  differentiations	
  in	
  robots’	
  genders	
  could	
  be	
  seen	
  in	
  the	
  graph	
  
above.	
   It	
   shows	
   very	
   close	
   results	
   between	
   the	
   female	
   and	
   male	
   forms	
   in	
   each	
   of	
   the	
  
alternatives.	
   This	
   means	
   that	
   we	
   don’t	
   actually	
   need	
   these	
   differentiations	
   between	
   the	
  
genders	
  and	
  we	
  could	
  reduce	
  the	
  number	
  of	
  alternatives	
  to	
  7,	
  instead	
  of	
  11.	
  	
  This	
  reduces	
  
the	
  time	
  needed	
  for	
  running	
  the	
  real	
  tests	
  as	
  well.	
  
Then	
  comes	
  the	
  question:	
  Which	
  gender	
  should	
  we	
  take	
  away?	
  
The	
  answer	
  comes	
  from	
  the	
  two	
  graphs	
  below	
  –	
  Figure	
  33	
  and	
  Figure	
  34.	
  	
   	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.49	
  /	
  56	
  
	
  
	
  
Table	
  1.	
  Data	
  results	
  from	
  the	
  pretests	
  
RHI:	
  Investigation	
  of	
  the	
  uncanny	
  valley	
   	
   Alessya	
  Ivanova	
  
	
  
p.50	
  /	
  56	
  
	
  
	
  
Figure	
  33.	
  The	
  uncanny	
  valley	
  graphs	
  for	
  both	
  genders	
  of	
  the	
  robots	
  
	
  
	
  
Figure	
  34.	
  Human	
  likeness	
  curves	
  for	
  both	
  participants’	
  genders	
  
	
  
The	
  Uncanny	
  valley	
  for	
  robot’s	
  genders	
  
diplomarbeit-alesiaivanova-clear
diplomarbeit-alesiaivanova-clear
diplomarbeit-alesiaivanova-clear
diplomarbeit-alesiaivanova-clear
diplomarbeit-alesiaivanova-clear
diplomarbeit-alesiaivanova-clear

Contenu connexe

En vedette

Creative concepts
Creative conceptsCreative concepts
Creative conceptsOlgaNazarko
 
DBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutów
DBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutówDBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutów
DBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutówDorian Szymański
 
Smartfony 2013-2015 (Connectmedica)
Smartfony 2013-2015 (Connectmedica)Smartfony 2013-2015 (Connectmedica)
Smartfony 2013-2015 (Connectmedica)Dorian Szymański
 
Determinan status gizi pada siswa sekolah dasar
Determinan status gizi pada siswa sekolah dasarDeterminan status gizi pada siswa sekolah dasar
Determinan status gizi pada siswa sekolah dasarFuadrizalfauzi
 

En vedette (9)

peluqueria
peluqueriapeluqueria
peluqueria
 
Reece
ReeceReece
Reece
 
Creative concepts
Creative conceptsCreative concepts
Creative concepts
 
DBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutów
DBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutówDBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutów
DBE Academy - szkolenia dla firm farmaceutycznych, lekarzy i farmaceutów
 
Task 1 scott
Task 1  scottTask 1  scott
Task 1 scott
 
Smartfony 2013-2015 (Connectmedica)
Smartfony 2013-2015 (Connectmedica)Smartfony 2013-2015 (Connectmedica)
Smartfony 2013-2015 (Connectmedica)
 
Determinan status gizi pada siswa sekolah dasar
Determinan status gizi pada siswa sekolah dasarDeterminan status gizi pada siswa sekolah dasar
Determinan status gizi pada siswa sekolah dasar
 
Asfiksia
AsfiksiaAsfiksia
Asfiksia
 
Christine Murphy Resume
Christine Murphy ResumeChristine Murphy Resume
Christine Murphy Resume
 

Similaire à diplomarbeit-alesiaivanova-clear

humanrobotinteraction-180527054931 (1) (1).pdf
humanrobotinteraction-180527054931 (1) (1).pdfhumanrobotinteraction-180527054931 (1) (1).pdf
humanrobotinteraction-180527054931 (1) (1).pdfHeenaSyed6
 
A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...
A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...
A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...Heather Strinden
 
Human–Robot Interaction: Status and Challenges. Sheridan MIT
Human–Robot Interaction: Status and Challenges. Sheridan MITHuman–Robot Interaction: Status and Challenges. Sheridan MIT
Human–Robot Interaction: Status and Challenges. Sheridan MITeraser Juan José Calderón
 
Ergonomics-for-One in a Robotic Shopping Cart for the Blind
Ergonomics-for-One in a Robotic Shopping Cart for the BlindErgonomics-for-One in a Robotic Shopping Cart for the Blind
Ergonomics-for-One in a Robotic Shopping Cart for the BlindVladimir Kulyukin
 
Design Validation and Analysis of a Humanoid Robot
Design Validation and Analysis of a Humanoid RobotDesign Validation and Analysis of a Humanoid Robot
Design Validation and Analysis of a Humanoid RobotIJSRD
 
A deep dive into the mechanics of human-robot interaction
A deep dive into the mechanics of human-robot interactionA deep dive into the mechanics of human-robot interaction
A deep dive into the mechanics of human-robot interactionReputationGuards2
 
Running head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docx
Running head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docxRunning head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docx
Running head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docxtoltonkendal
 
Wireless Pick & Place Robot
Wireless Pick & Place RobotWireless Pick & Place Robot
Wireless Pick & Place RobotMarmik Kothari
 
Humanoid project report
Humanoid project reportHumanoid project report
Humanoid project reportMayank Rai
 
Human Robot Interaction
Human Robot InteractionHuman Robot Interaction
Human Robot Interactionijtsrd
 
Vinod Robotics
Vinod RoboticsVinod Robotics
Vinod RoboticsColloquium
 
The study of attention estimation for child-robot interaction scenarios
The study of attention estimation for child-robot interaction scenariosThe study of attention estimation for child-robot interaction scenarios
The study of attention estimation for child-robot interaction scenariosjournalBEEI
 
Robots a threat_or_an_opportunity_m2_deipm
Robots a threat_or_an_opportunity_m2_deipmRobots a threat_or_an_opportunity_m2_deipm
Robots a threat_or_an_opportunity_m2_deipmAnastasia Romanski
 
Николаос Мавридис. От Интерактивных роботов к Человеку-машинному облаку
Николаос Мавридис. От Интерактивных роботов к Человеку-машинному облакуНиколаос Мавридис. От Интерактивных роботов к Человеку-машинному облаку
Николаос Мавридис. От Интерактивных роботов к Человеку-машинному облакуSkolkovo Robotics Center
 

Similaire à diplomarbeit-alesiaivanova-clear (20)

humanrobotinteraction-180527054931 (1) (1).pdf
humanrobotinteraction-180527054931 (1) (1).pdfhumanrobotinteraction-180527054931 (1) (1).pdf
humanrobotinteraction-180527054931 (1) (1).pdf
 
A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...
A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...
A Methodological Variation For Acceptance Evaluation Of Human-Robot Interacti...
 
Human–Robot Interaction: Status and Challenges. Sheridan MIT
Human–Robot Interaction: Status and Challenges. Sheridan MITHuman–Robot Interaction: Status and Challenges. Sheridan MIT
Human–Robot Interaction: Status and Challenges. Sheridan MIT
 
Robotics report
Robotics reportRobotics report
Robotics report
 
Ergonomics-for-One in a Robotic Shopping Cart for the Blind
Ergonomics-for-One in a Robotic Shopping Cart for the BlindErgonomics-for-One in a Robotic Shopping Cart for the Blind
Ergonomics-for-One in a Robotic Shopping Cart for the Blind
 
Design Validation and Analysis of a Humanoid Robot
Design Validation and Analysis of a Humanoid RobotDesign Validation and Analysis of a Humanoid Robot
Design Validation and Analysis of a Humanoid Robot
 
A deep dive into the mechanics of human-robot interaction
A deep dive into the mechanics of human-robot interactionA deep dive into the mechanics of human-robot interaction
A deep dive into the mechanics of human-robot interaction
 
Running head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docx
Running head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docxRunning head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docx
Running head ROBOTIC SURGERY TECHNOLOGY1OPERATING SYSTE.docx
 
Wireless Pick & Place Robot
Wireless Pick & Place RobotWireless Pick & Place Robot
Wireless Pick & Place Robot
 
ICSR15 Paper
ICSR15 PaperICSR15 Paper
ICSR15 Paper
 
Humanoid project report
Humanoid project reportHumanoid project report
Humanoid project report
 
Human Robot Interaction
Human Robot InteractionHuman Robot Interaction
Human Robot Interaction
 
Vinod Robotics
Vinod RoboticsVinod Robotics
Vinod Robotics
 
The study of attention estimation for child-robot interaction scenarios
The study of attention estimation for child-robot interaction scenariosThe study of attention estimation for child-robot interaction scenarios
The study of attention estimation for child-robot interaction scenarios
 
E0352435
E0352435E0352435
E0352435
 
Robots a threat_or_an_opportunity_m2_deipm
Robots a threat_or_an_opportunity_m2_deipmRobots a threat_or_an_opportunity_m2_deipm
Robots a threat_or_an_opportunity_m2_deipm
 
Николаос Мавридис. От Интерактивных роботов к Человеку-машинному облаку
Николаос Мавридис. От Интерактивных роботов к Человеку-машинному облакуНиколаос Мавридис. От Интерактивных роботов к Человеку-машинному облаку
Николаос Мавридис. От Интерактивных роботов к Человеку-машинному облаку
 
Ar35247251
Ar35247251Ar35247251
Ar35247251
 
Ts 2 b topic
Ts 2 b topicTs 2 b topic
Ts 2 b topic
 
IEEE-ROBIO-2013-115-120
IEEE-ROBIO-2013-115-120IEEE-ROBIO-2013-115-120
IEEE-ROBIO-2013-115-120
 

diplomarbeit-alesiaivanova-clear

  • 1.     Lehrstuhl und Institut für Arbeitswissenschaft der RWTH Aachen Direktor: Univ.-Prof. Dr.-Ing. Dipl.-Wirt.-Ing. Christopher Marc Schlick Telefon: 0241 80-99 440 Telefax: 0241 80-92 131 info@iaw.rwth-aachen.de www.iaw.rwth-aachen.de Dienstgebäude: Bergdriesch 27 – D-52062 Aachen Postanschrift: D-52056 Aachen                   Master  Thesis Robot-human interaction: Investigation of the uncanny valley using different designed robot alternatives Angefertigt von cand.ing. Alesia St. Ivanova Matr.-Nr.: 328510 Betreuer: Univ.-Prof. Dr.-Ing. Dipl.-Wirt.-Ing. Christopher M. Schlick Mitbetreuen der wissenschaftl. Mitarbeiter: Dipl.-Ing. Christopher Brandl Aachen, den 22.03.2013    
  • 2. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.2  /  56     Contents     1. Introduction  ___________________________________________________________    p.3   2. Research  ______________________________________________________________    p.5   1) Previous  studies  and  service  robots   i. Care-­‐o-­‐bot   ii. PAMM   iii. RIBA   iv. Taizo   v. Exoskeletons   2) The  Uncanny  valley   3) Performance  assessment   3. The  aim  ______________________________________________________________    p.23   4. Design  concept  ________________________________________________________    p.23   5. Design  project  _________________________________________________________    p.25   6. Pretest  design  _________________________________________________________    p.33   1) Participant  and  procedure   2) Visualizations     7. Pretest  results  and  analyses  ______________________________________________    p.48   8. Conclusions  ___________________________________________________________    p.53   9. References  ____________________________________________________________    p.55            
  • 3. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.3  /  56     Introduction       More  than  30%  of  the  medical  expenses  all  over  the  world  can  be  related  to  elderly  people.   And  these  expenses  are  increasing  drastically  nowadays.  By  the  year  2030  the  number  of  60-­‐ year-­‐old   people   will   have   doubled.   Wheelchair   dependents   have   difficulty   moving   from   a   seat,  to  their  wheelchair  and  back  without  a  caregivers  help  or  other  lift  mechanisms.  [1]     While  1  in  3  nurses  are  expected  to  develop  back  injuries  while  moving  and  lifting  patients   and  50%  of  non-­‐ambulatory  patients  fall  to  the  floor,  all  aspects  of  Service  robots  (e.g.  the   HLPR   Chair)   provide   for   independent   patient   mobility   especially   on   lifting   and   placing   patients  to  eliminate  or  significantly  reduce  this  back  injury  issue.   With  fewer  caregivers  and  more  elderly,  there  is  a  need  for  improving  these  technical  aids   for  providing  independent  assistance.  [2]  It  is  envisioned  that  in  the  near  future  personal   mobile  robots  will  be  assisting  people  in  their  daily  lives.  An  essential  characteristic  shaping   the  design  of  personal  robots  is  the  fact  that  they  must  be  accepted  by  human  users.   In   general,   it   seems   natural   to   assume   that   the   more   human   looking   the   robots   are,   the   more  likely  they  are  to  provoke  the  usual  responses  people  show  to  each  other.  However,   even  subtle  flaws  in  appearance  and  movement  seem  strange  and  eerie  in  very  humanlike   robots.           This  paper  explores  the  acceptance  of  mobile  personal  service  robots,  by  focusing  on  the   psychological  effects  of  robot  appearance.  The  level  of  comfort  the  robot  causes  to  human   subjects  is  analyzed  according  to  the  effects  of  robot  design.  The  information  gained  from   surveys  taken  by  40  to  60  human  subjects  can  be  used  to  obtain  a  better  understanding  of   what  characteristics  make  up  personal  robot  appearances  that  are  most  acceptable  to  the   human   users.   The   results   gained   from   this   study   should   yield   useful   insights   into   how   to   calibrate  robot  appearance  so  that  users  of  service  robots  in  future  will  be  less  disaffected   due  to  design  feature  limitations  which  do  not  meet  their  initial  expectations.       This   study   includes   a   research   about   service   robot   design,   users’   preferences   and   the   influence   of   robot   appearance.   An   overview   of   the   Mori’s   “Uncanny   valley”   hypothesis   is   introduced,   as   well.   Furthermore,   design   alternatives   development   is   presented.   Finally,   pretest  results  and  conclusion  for  further  studies  are  introduced.                          
  • 4. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.4  /  56     Problem  definition     The   empirical   studies   investigating   the   importance   of   the   “uncanny   valley”   all   have   in   common  the  low  number  of  factor  levels  of  robot  appearances  or  the  equidistance  in  the   intervals  between  them.   For  example  in  [3]  they  used  only  three  “Peoplebot”  robot  versions  in  their  video-­‐based  HRI   trials  to  support  a  portion  of  the  left  hand  side  of  Mori’s  theoretically  proposed  “uncanny   valley”  diagram.     In  [4]  only  highly  realistic  robots  were  used  to  empirically  test  the  right  hand  side  of  the   hypothesis   by   the   following   statements:   a)   highly   realistic   robots   are   liked   less   than   real   humans  and  b)  the  highly  realistic  robot’s  movement  decreases  its  likeability.   According  to  MacDorman’s  analyses  on  Mori’s  theory  for  the  uncanny  valley  it  is  “possible  to   produce  a  safe  familiarity  by  a  non-­‐humanlike  design”,  concerning  the  robot  appearance  and   its  acceptance.   More  prolonged  experiments,  using  finer  gradations  of  robot  appearances  and  behavior  are   required  in  order  to  give  more  data  sample  points  from  the  diagram  and  to  provide  more   extensive  evidence,  which  can  then  be  used  to  refine  the  parameters  which  define  human   perception  of  robot  appearance  and  how  these  can  be  applied  to  developing  principles  for   robot  aesthetics  in  different  user  environments.            
  • 5. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.5  /  56     Research     In   this   section   several   service   robots   are   presented,   some   of   them   already   in   use,   while   others   still   in   a   development   phase.     Also   experiments   testing   the   social   acceptance   of   robots   with   varied   human-­‐likeness,   the   importance   of   tentative   acceptance   properties   together  with  the  results  from  these  studies,  are  described.     Some   of   the   researched   studies   explore   the   interactions   between   humans   and   mobile   personal  robots,  by  focusing  on  the  psychological  effects  of  robot  behavior  patterns  during   task  performance.  The  aim  of  the  research  is  to  gain  some  guidelines  for  robot  design  from   past  studies  in  this  area.       1. Care-­‐O-­‐bot       Care-­‐O-­‐bot  was  designed  and  implemented  by  Fraunhofer  IPA,  Stuttgart.  The  Care-­‐O-­‐bot  is  a   mobile  service  robot,  which  has  the  capability  to  perform  fetch  and  carry  and  various  other   supporting   tasks   in   home   environments.   Main   emphasis   is   laid   on   integrating   communicational  and  social  features,  like  video  telephone,  automatic  emergency  calls  and   other  interactive  communication  (Figure  1).       Figure  1.  Care-­‐o-­‐bot®3  in  action  [1]   Care-­‐O-­‐bot®  3 was  released  relatively  soon  in  2008,  after  over  10  years  of  development,  and   excels  in  its  user-­‐interaction  oriented  design.  Nevertheless,  it  is  equipped  with  leading  edge   technology,  which  is  highly  integrated  into  a  very  compact  form.  This  convergence  of  design   and  technology  accounts  for  the  idea  of  Care-­‐O-­‐bot®  3  being  a  product  vision  for  a  robot   butler,  combining  technological  aspects  with  a  compact  and  user  friendly  design.   The  primary  interface  between  Care-­‐O-­‐bot®  3 and  the  user  consists  of  a  tray  attached  to  the   front  of  the  robot,  which  carries  objects  for  exchange  between  the  human  and  the  robot.   The   tray   includes   a   touch   screen   and   retracts   automatically   when   not   in   use.   The   basic  
  • 6. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.6  /  56     concept  developed  was  to  define  two  sides  of  the  robot:  “Working  side”  and  “Serving  side”.   The  concept  behind  using  the  tray  to  interact  with  the  user  is  to  reduce  possible  users’  fears   of  mechanical  parts  by  having  smooth  surfaces  and  a  likable  appearance  (Figure  1.a).  On  the   technical  side,  it  is  much  easier  to  ensure  collision  free  interaction  with  the  static  tray  than   with  a  robotic  arm  moving  freely  in  3-­‐D-­‐space.  Using  these  described  interaction  concept,   Care-­‐O-­‐bot®3   enables   the   safe   executing   of   fetch   and   carry   tasks   and   thus   provides   the   potential  to  operate  a  mobile,  manipulating  robot  safely  in  public  environments.     Care-­‐O-­‐bot®  3  is  driven  by  four  wheels.  Each  wheel’s  orientation  and  rotational  speed  can  be   set  individually.  The  wheeled  drive  was  preferred  to  legged  locomotion  because  of  safety  (no   risk  of  falling)  and  stability  during  manipulation.  [5]       Figure  1.a).  Care-­‐o-­‐bot®3  “two-­‐sides”  concept  [5]       2. PAMM       PAMM   (Personal   Aid   for   Mobility   and   Monitoring)   is   intended   to   assist   the   elderly   living   independently   or   in   senior   Assisted   Living   Facilities.   It   provides   physical   support   and   guidance,  and  it  monitors  the  user's  basic  vital  signs.     Figure  2  summarizes  the  PAMM  Concept  design  [6].       For   models   released   in   the   early   21st   century   their   appearance   was   closely   to   machine-­‐ looking,  which  is  very  simplistic  and  does  not  intend  to  impress  with  any  modern  or  futuristic   body  (Figure.2.a),  while  the  latest  Walking  aid  devices  impress  with  smoother  shapes  and   nice  look,  associated  with  a  personal  vehicle  (Figure  2.b).  Their  newer  technology  of  driving   allows  using  only  2  wheels  instead  of  3  or  4.    
  • 7. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.7  /  56         Figure  2.  PAMM  System  Concept           Figure  2.a).  PAMM  devices  from  2000’s  [7]    
  • 8. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.8  /  56         Figure  2.b).  PAMM  devices  from  Toyota,  2008  [8]         3. RIBA     RIBA  (Robot  for  Interactive  Body  Assistance)  is  said  to  be  the  first  robot  that  can  lift  up  or  set   down  a  real  human  (up  to  61kg/134lbs)  from  or  to  a  bed  or  wheelchair.  RIBA  does  this  using   a   combination   of   its   very   strong   human-­‐like   arms   and   by   novel   tactile   guidance   methods   using   high-­‐accuracy   tactile   sensors.   RIBA   was   developed   by   integrating   RIKEN's   control,   sensor,  and  information  processing  and  TRI's  material  and  structural  design  technologies.     It  might  look  like  a  cross  between  a  snowman  and  a  badly-­‐designed  toy  polar  bear,  but  the   nursing  fraternity  should  appreciate  this  robot  that  can  lift  patients  (Figure.3)  in  and  out  of   beds   and   wheelchairs   on   command,   while   at   the   same   time   saving   nurses’   backs   and   improving  patient  care  and  safety.  [9]   The  robot’s  body  is  covered  with  soft  materials  and  the  elbow  and  waist  joints  are  isolated,   making   RIBA   safe   for   physical   interactions   with   humans.   This   softness   also   contributes   to   patient  comfort  when  they  are  being  lifted.  A  teddy  bear  shape  was  deliberately  used  to  put   patients  at  ease  and  to  give  a  friendly,  non-­‐threatening,  appearance.  
  • 9. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.9  /  56       Figure  3.  Photograph  of  RIBA-­‐I  and  RIBA-­‐II  lifting  a  patient     4. Taizo   The   National   Institute   of   Advanced   Industrial   Science   and   Technology   (AIST)   and   Ibaraki   Prefectural   Health   Plaza   in   Japan   are   developping   ‘Taizo’,   a   humanoid   robot   designed   to   lead  the  elderly  in  physical  exercises.Taizo  (Figure  4.),  which  is  a  play  on  the  word  “taisou”   meaning  “calisthenics”,  stands  72cm  tall  and  is  dressed  in  a  velvety  space  suit.  He  sports  a   clown-­‐like  grin  that  is  supposed  to  look  silly  to  put  the  older  generation,  who  are  often  a   little  frightened  by  new  technology,  at  ease.  [10]         Figure  4.  Taizo    
  • 10. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.10  /  56       5. Exoskeletons     An  exoskeleton  is  a  distinctive  kind  of  robot  to  be  worn  as  an  overall,  effectively  supporting   or,   in   some   cases   substituting   for,   the   user’s   own   movements.   The   development   of   exoskeletons   can   lead   to   important   changes   in   the   rehabilitation   of   disabled   people   by   introducing   an   alternative   to   wheelchairs.   Exoskeletons   can   be   an   efficient   tool   in   the   restoration  of  upper  limb  functions,  and  they  can  support  therapists  and  caregivers  in  tasks   that  require  major  physical  effort.  The  functionality  of  exoskeleton  can  easily  be  extended  by   a  “disabled  person  integrated  IT  environment”,  described  by  authors.  Exoskeletons  can  also   be  adapted  to  the  needs  of  severely  ill  or  aged  people.     Exoskeletons  can  be  divided  into  two  categories:  those  for  all  four  extremities  (arms/legs)   and   those   for   the   lower   extremities   only   (Figure   5.).   Exoskeletons   are   controlled   by   the   user’s  movements  and  do  not  need  any  external  control  terminal  (with  the  exception  of  a   service   terminal).     The   main   parts   are:   the   frame;   the   power   system,   including   engines,   actuators  and  batteries;  and  the  control  system  with  sensors.     Figure  5.  Examples  of  medical  exoskeletons:     A)  HAL5  –  version  for  four  extremities;  B)  ReWalk        
  • 11. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.11  /  56       Figure  6.  Robot  suits  HAL-­‐5   The   robotics   geeks   at   Honda   have   developed   an   exoskeleton   that   is   worn   like   shoes   to   support  the  body  and  protect  the  joints,  something  the  automaker  says  could  reduce  injuries   on  assembly  lines  but  also  might  help  the  elderly  get  around  more  easily.       The  device  resembles  a  bicycle  seat  joined  to  a  pair  of  shoes  and  fits  between  the  legs  to   help  the  user  walk,  crouch  and  stand  without  excessive  stress  on  the  hips,  knees  and  ankles.   Honda   is   testing   the   "walking   assist   device"   (Figure   7.b)   at   a   vehicle   assembly   line   in   Sayama,  Japan,  and  says  robo-­‐legs  could  help  anyone  who  spends  a  lot  of  time  on  their  feet.   More  than  that,  it  could  help  the  elderly  and  infirm  by  making  it  easier  to  get  around.   In  contrast  to  the  complexity  of  the  HAL  (Figure  6.),  the  Honda  devices’  simplicity  may  be   their  strength  (Figure  7.b.)   For  example,  U3-­‐X  personal  mobility  prototype  with  its  compact  size  and  one-­‐wheel-­‐drive   personal  mobility  prototype  was  designed  to  be  friendly  to  the  user  and  people  around  it  by   making  it  easier  for  the  rider  to  reach  the  ground  from  the  footrest  and  placing  the  rider  on   roughly  the  same  eye  level  as  other  people  or  pedestrians.  [11]     Figure  7.a.  Left  to  right:  U3-­‐X  personal  mobility  prototype,  Bodyweight  Support  Assist Device,  Stride   Management  Support  Assist  Device,  ASIMO  humanoid  robot  
  • 12. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.12  /  56     Figure  7.b.  Robo-­‐legs  from  HONDA  [12]     Tendencies  in  service  robot  design     All   of   the   researched   robots   above   show   today’s   tendencies   in   human-­‐robot   interaction   design  –  a  combination  of  high  technological  execution,  smooth  shapes  and  user  friendly   interface,  materials  depend  on  their  purpose  and  use  in  daily  life.  For  the  current  study  very   important   guidelines   can   be   gained   from   medical   purpose   robots   such   as   Riba   or   Taizo,   because  they  were  designed  in  order  to  make  people  feel  comfortable  and  friendly  in  their   companion.  This  would  help  us  to  choose  a  similar  approach  in  designing  a  head  and/or  a   face  for  our  experimental  robots  and  investigate  this  part  of  the  uncanny  valley  graph  which   is  more  human-­‐like.     The  other  devices  like  PAMM,  Care-­‐o-­‐bot  and  exoskeletons  are  shown  especially  for  their   driving  systems  –  how  do  they  execute  movements,  tasks  and  how  they  approach  to  users.   Obviously  the  more  human-­‐like  is  the  robot,  the  more  complicated  “walking”  system  is.  The   machine-­‐looking,   such   as   PAMM   and   Care-­‐o-­‐bot   use   wheels,   and   even   more   machine-­‐ looking  one  would  have  chains  like  in  a  tank.  This  kind  of  machine  parts  make  the  robots   more  maneuverable  and  smooth  in  movements,  while  walking-­‐resembling  robots  are  still  in   development  for  improving  realistic  movements.            
  • 13. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.13  /  56     The  Uncanny  valley      “Climbing  a  mountain  is  an  example  of  a  function  that  does  not  increase  continuously:  a  person's   altitude  does  not  always  increase  as  the  distance  from  the  summit  decreases  owing  to  the  intervening   hills  and  valleys.  I  have  noticed  that,  as  robots  appear  more  humanlike,  our  sense  of  their  familiarity   increases  until  we  come  to  a  valley.  I  call  this  relation  the  “uncanny  valley”.”   Masahiro  Mori     Recently  prosthetic  hands  have  improved  greatly,  and  we  cannot  distinguish  them  from  real   hands   at   a   glance.   Some   prosthetic   hands   attempt   to   simulate   veins,   muscles,   tendons,   finger  nails,  and  finger  prints,  and  their  color  resembles  human  pigmentation.  But  this  kind   of  prosthetic  hand  is  too  real  and  when  one  notices  it  is  prosthetic,  they  have  a  sense  of   strangeness.  So  if  we  shake  the  hand,  we  are  surprised  by  the  lack  of  soft  tissue  and  cold   temperature.   In   this   case,   there   is   no   longer   a   sense   of   familiarity.   It   is   uncanny.   In   mathematical   terms,   strangeness   can   be   represented   by   negative   familiarity,   so   the   prosthetic   hand   is   at   the   bottom   of   the   valley.   So   in   this   case,   the   appearance   is   quite   human  like,  but  the  familiarity  is  negative.  This  is  the  uncanny  valley.       Figure  8.  The  Uncanny  Valley  graph     As  a  robot  designer,  Mori  graphed  what  he  saw  as  the  relation  between  human  likeness  and   perceived  familiarity:  familiarity  increases  with  human  likeness  until  a  point  is  reached  at   which   deviations   from   human   appearance   and   behavior   create   unnerving   effect.   This   he   called  the  uncanny  valley.  According  to  Mori,  movement  amplifies  the  effect  (Figure  8.)     In  the  World  Expo  held  in  Osaka  2009  were  displayed  robots  with  a  very  elaborate  design.   For   example,   one   robot   has   29   artificial   muscles   in   the   face   to   make   humanlike   facial   expressions.  According  to  the  designer,  laughing  is  a  kind  of  sequence  of  face  distortions,   and  the  distortion  speed  is  an  important  factor.  If  the  speed  is  cut  in  half,  laughing  looks   unnatural.  This  illustrates  how  slight  variations  in  movement  can  cause  a  robot,  puppet,  or   prosthetic  hand  to  tumble  down  into  the  uncanny  valley.  
  • 14. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.14  /  56     The  author  hopes  to  design  robots  or  prosthetic  hands  that  will  not  fall  into  the  uncanny   valley.   So   he   recommends   designers   to   take   the   first   peak   as   a   goal   in   building   robots   rather  than  the  second.  Although  the  second  peak  is  higher,  there  is  a  far  greater  risk  of   falling  into  the  uncanny  valley.  They  predict  that  it  is  possible  to  produce  a  safe  familiarity   by  a  nonhumanlike  design.  This  is  a  good  point  to  take  into  a  consideration.       A   good   example   is   glasses.   Glasses   do   not   resemble   the   real   eyeball,   but   this   design   is   adequate  and  can  make  the  eyes  more  charming.  So  designers  should  follow  this  principle   when  design  prosthetic  eyes.  An  elegant  prosthetic  hand  can  be  created  -­‐  one  that  must  be   fashionable.  Artist  who  makes  statues  of  Buddha  created  a  model  of  a  human  hand  that  is   made   from   wood.   The   fingers   bend   at   their   joints.   The   hand   has   no   finger   print,   and   it   assumes  the  natural  color  of  wood.  But  it  still  looks  beautiful  and  there  is  no  sense  of  the   uncanny.  [13]   Christoph  Bartneck,  Takayuki  Kanda,  Hiroshi  Ishiguro,  and  Norihiro  Hagita  (Members  of  IEEE)   conducted  a  study  which  attempted  to  empirically  test  two  aspects  of  Mori’s  hypothesis.   First,  they  were  interested  in  the  degree  to  which  highly  realistic  androids  were  perceived   differently  from  a  human.  The  uncanny  valley  hypothesis  predicts  that  androids  would  be   perceived  as  less  human-­‐like  and  less  likeable  compared  to  humans.     To  test  this  hypothesis,  they  used  Hiroshi  Ishiguro  and  his  robotic  copy  named  “Geminoid  HI-­‐ 1”   (Figure   9).   Also,   they   wanted   to   test   whether   a   more   humanlike   android   would   be   perceived  as  more  likeable  compared  to  a  less  human-­‐like  robot.  Accordingly,  they  made  a   small  alteration  to  Geminoid  HI-­‐1  to  make  it  appear  less  humanlike.       Figure  9.  Hiroshi  Ishiguro  and  his  robotic  double  Geminoid  HI-­‐1     Second,  they  were  interested  in  the  effect  of  the  android’s  movement.  Mori’s  hypothesis   predicts  that  movement  intensifies  the  users’  perception  of  an  android.  A  moving  android   would  be  perceived  differently  from  an  inert  android.          
  • 15. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.15  /  56     In  summary,  the  following  three  hypotheses  were  interesting  for  the  study:     1.  Androids  that  are  distinguishable  from  humans  will  be  liked  less  than  humans.   2.  A  fully  moving  android  will  be  liked  differently  compared  to  an  android  that  is  limited  in  its  movements.   3.  Androids  with  different  levels  of  anthropomorphism  will  be  liked  differently.     They   conducted   a   3   (anthropomorphism)   x   2   (movement)   experiment   in   which   anthropomorphism   was   the   within   participants   factor   and   movement   was   the   between   participants   factor.   The   anthropomorphism   factor   had   three   conditions:   masked   android,   android,  and  human.  The  movement  factor  had  two  conditions:  full  movement  and  limited   movement.     The   participants   in   this   study   were   19   men   and   13   women   in   their   early   20’s   attending   Japanese  universities  in  the  Kansai  area.  The  male  and  female  participants  were  distributed   approximately   even   across   the   experimental   conditions.   They   were   not   exposed   to   any   previous  study  in  the  laboratory.  This  study  was  conducted  shortly  before  the  official  release   of  the  Geminoid  HI-­‐1  and  hence  they  were  not  exposed  to  the  considerable  media  exposure   that  the  Geminoid  HI-­‐1  received.   The  participants  were  welcomed  and  then  asked  to  fill  in  a  questionnaire.  They  were  seated   on   a   chair   that   was   placed   one   meter   away   from   the   android/person.   Afterwards   the   experimenter  introduced  them  to  each  other  without  explicitly  labeling  Ishiguro  as  a  human   and  the  androids  as  robots.     Conclusions  from  the  study:  Against  Mori’s  prediction,  androids  that  were  distinguishable   from   humans   were   not   liked   less   than   humans.   The   results   showed   that   the   participants   were  able  to  distinguish  between  the  human  stimulus  and  the  android  stimuli.  The  human   was  rated  as  being  significantly  more  human-­‐like  compared  to  the  two  androids.  However,   the  ratings  for  likeability  were  not  significantly  different.  This  result  does  not  support  Mori’s   hypothesis.  Two  possible  interpretations  could  be  possible.  On  the  one  hand,  there  really   could  be  no  difference  between  the  likeability  of  humans  and  that  of  androids.  On  the  other   hand,  likeability  could  be  a  more  complex  phenomenon.  They  speculate  that  the  participants   might   have   used   different   standards   to   evaluate   the   likeability   of   the   human   and   the   androids.  As  a  robot,  the  displayed  android  might  have  been  likeable  to  the  same  degree  as   the  human  was  likeable  as  a  human;  however,  the  expectations  for  these  two  categories   might  have  been  different.     The  results  of  this  study  cannot  confirm  Mori’s  hypothesis  of  the  Uncanny  Valley.  The  robots’   movements  and  their  level  of  anthropomorphism  may  be  complex  phenomena  that  cannot   be   reduced   to   two   factors.   Movement   contains   social   meanings   that   may   have   direct   influence  on  the  likeability  of  a  robot.  The  robot’s  level  of  anthropomorphism  does  not  only   depend   on   its   appearance   but   also   on   its   behavior.   A   mechanical-­‐looking   robot   with   appropriate  social  behavior  can  be  anthropomorphized  for  different  reasons  than  a  highly   humanlike  android.  Again,  Mori’s  hypothesis  appears  to  be  too  simplistic.  [4]    
  • 16. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.16  /  56       Figure  10.  Graph  showing  results  of  MacDorman  and  Hiroshi  Ishiguro  study  on  the  Uncanny  valley     Another  study  from  MacDorman  and  Ishiguro  is  presented  here.  The  average  ratings  of  45   Indonesian  participants  on  scales  of  human  likeness,  familiarity,  and  eeriness  are  presented   for  the  above  figure  (Figure  10).  The  images  morph  from  a  mechanical-­‐looking  humanoid  on   the  left  to  an  android  in  the  center  to  a  human  being  on  the  right.     For  the  given  images,  they  reveal  an  uncanny  region,  both  on  the  strange-­‐familiar  scale  and   on   the   eeriness   scale.   In   the   follow-­‐up   discussion   David   Han-­‐son   argued   that   a   valley   of   eeriness  was  not  inevitable  for  a  specific  range  of  human  likeness.  He  claims  that,  across  the  
  • 17. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.17  /  56     spectrum   of   human   likeness,   it   is   possible   to   design   androids   that   are   not   uncanny.   In   a   follow-­‐up   experiment   in   which   intermediate   images   were   designed,   adapting   more   attractive,  cartoon-­‐like  features,  rather  than  simply  morphed,  Hanson  eliminates  the  valley   from   his   results.   In   addition,   Hanson   notes   that   very   abstract   robots   and   cosmetically   atypical  people  can  be  uncanny,  although  they  are  far  from  the  posited  region  of  the  valley   in  terms  of  human  likeness.       Do  Hanson’s  results  mean  that  the  uncanny  valley  does  not  exist?  They  may  suggest  that  the   uncanny  valley  is  not  inevitable  or  that  designers  with  finesse  can  moderate  it  in  situations   that   involve   still   images.   Nevertheless,   human   beings   do   seem   to   be   highly   sensitive   to   imperfections  in  near  humanlike  robots,  both  in  their  looks  and  movements,  which  is  why   androids  are  potentially  very  useful  in  studying  human  perception.  Furthermore,  only  limited   conclusions  can  be  drawn  from  ratings  of  still  images,  which  are  static,  modern  inventions   appearing   after   human   beings   evolved.   For   instance,   cartoon   images   can   be   aesthetically   pleasing,  but  if  real  people  could  exist  with  the  same  proportions,  they  would  be  considered   freaks.  There  is  no  way  to  evaluate  whether  a  still  image  is  responding  as  predicted,  because   they  cannot  respond  at  all.  [14]          
  • 18. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.18  /  56     Performance  assessment       The  effectiveness  of  service  robots  cannot  be  assessed  only  by  performance  criteria  typically   found   for   industrial   robots.   The   performance   criteria   of   service   robots   lie   within   the   satisfaction   of   their   users.   Therefore,   it   is   necessary   to   measure   the   users’   perception   of   service   robots,   since   these   cannot   be   measured   within   the   robots   themselves.   Below   are   some  of  the  main  criteria  used  in  these  measurements.     Social  acceptance   What   can   be   considered   to   be   the   important   criteria   for   social   acceptance   of   robots   in   homes  and  health  care?  It  will  most  likely  vary  with  both  application  area  and  culture.  One   example  of  such  criteria  might  be  the  size  of  the  robot,  where  several  aspects  have  been   reported.   One   such   consideration   states   that   the   robot   has   to   be   smaller   than   human   in   order  not  to  "dominate"  the  human  user.  This  criteria  (size)  can  be  developed  further  in  the   context   of   home   care   for   handicapped   people,   where   it   is   stated   that   the   robot   has   to   defend  its  space  though  its  functionality,  since  a  large  appliance  will  compete  with  the  other   support  equipment  that  is  already  in  the  care  environment  (such  as  wheel  chairs,  respiratory   equipment  etc.).  If  the  robot  size  is  too  big,  it  will  probably  simply  not  be  used.   Examples   of   similar   properties   that   will   be   important   to   look   at   in   the   social   acceptance   perspective  are  technical  issues  such  as:   • Weight  -­‐  the  impression  of  moving  weight   • Speed  -­‐  the  speed  of  the  robot’s  movement   • Agility  -­‐  the  speed  of  limb  movements   • Reliability  -­‐  the  functional  stability  of  the  robot   • There  are  also  more  emotional  and  possibly  culture  dependent  issues  such  as:   • Anthropomorphism  -­‐  should  the  robot  look  like  a  human  or  a  machine?   • Social  behavior  -­‐  which  behavior  characteristics  are  important  for  the  acceptance  of  the  robot?   • Autonomy  -­‐  should  the  robot  act  with  or  without  user  involvement?   • Distance  -­‐  how  big  should  be  the  minimum  distance  between  human  and  robot  (something   which  clearly  is  culturally  dependent)?   • Safety  and  security  -­‐  how  safe  does  a  person  feel  in  relation  to  the  robot?   • Reliability  -­‐  can  you  trust  the  robot  to  perform  the  tasks  and  do  this  in  the  right  way?     Taken  together,  these  variables  will  affect  the  user  in  one  way  or  another.  It  is  therefore   necessary   for   the   developers   and   designers   to   find   out   which   of   these   aspects   that   are   important  to  consider,  and  second,  how  the  properties  affect  the  social  acceptance  of  the   robot.  Below  some  of  them  are  described  in  more  details.     Anthropophormism   Anthropomorphism   refers   to   the   attribution   of   a   human   form,   human   characteristics,   or   human  behavior  to  nonhuman  things  such  as  robots,  computers,  and  animals.       Even  if  it  is  not  the  intention  of  the  design  of  a  certain  robot  to  be  as  humanlike  as  possible,   it   still   remains   important   to   match   the   appearance   of   the   robot   with   its   abilities.   A   too   anthropomorphic  appearance  can  evoke  expectations  that  the  robot  might  not  be  able  to   fulfill.  If,  for  example,  the  robot  has  a  human-­‐shaped  face  then  the  naive  user  will  expect  
  • 19. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.19  /  56     that  the  robot  is  able  to  listen  and  to  talk.  To  prevent  disappointment  it  is  necessary  for  all   developers  to  pay  close  attention  to  the  anthropomorphism  level  of  their  robots.     Anthropomorphism  is  a  constant  pattern  in  human  cognition  and  the  interaction  of  a  human   with  a  robot  (or  any  kind  of  machine)  cannot  completely  avoid  it.  According  to  Mori,  the  so-­‐ called  uncanny  valley  would  suggest  to  either  stay  in  the  area  of  very  non-­‐human,  toy-­‐like   robots,  or  to  create  a  robot  that  appears  to  be  almost  perfectly  human-­‐like,  because  a  robot   in  between  may  provoke  rather  fearful  responses.  [15].  Human-­‐like  appearance  is  likely  to   trigger  expectations  that  go  beyond  the  capabilities  of  a  machine.  But  being  humanoid  in   appearance  does  hardly  suffice  to  meet  the  expectancy  of  humanlike  reactions.  [16]     Likeability   There   is   a   growing   body   of   research   which   indicates   that   people   often   make   important   judgments  within  seconds  of  meeting  another  person,  sometimes  quite  unaware  of  both  the   obvious   and   subtle   signs   that   may   be   influencing   their   judgments.   Since   computers,   and   thereby  robots  in  particular,  are  to  some  degree  treated  as  social  actors,  it  can  be  assumed   that  people  are  able  to  judge  robots  in  a  similar  way.  Jennifer  Monahan  complemented  her   “liking”   question   with   5-­‐point   semantic   differential   scales:   nice/awful,   friendly/unfriendly,   kind/unkind,   and   pleasant/unpleasant,   because   these   judgments   tend   to   demonstrate   considerable   variance   in   common   with   “liking”   judgments.   Monahan   later   eliminated   the   kind-­‐unkind   and   pleasant-­‐unpleasant   items   in   her   own   analysis   since   they   did   not   load   sufficiently   in   a   factor   analysis   that   also   included   items   from   three   other   factors.   The   Cronbach’s   Alpha   of   0.68   therefore   relates   only   to   this   reduced   scale.   She   also   included   concepts   of   physical   attraction,   conversational   skills,   and   other   orientations,   which   might   become   an   element   of   the   questionnaire   series.   In   particular,   physical   attraction   might   require  additional  conceptual  and  social  consideration,  since  it  may  also  entail  sexuality.       Perceived  Safety   Safety  is  a  key  issue  for  robots  interacting  with  humans.  The  issue  has  received  considerable   attention  in  the  robotics  literature,  both  in  systems  and  standards  established  for  industrial   robots  and  for  service  robots  intended  for  use  in  the  home.  Examples  of  design-­‐concerning   category  of  safety  is  the  mechanical  redesign  which  includes  using  a  whole-­‐body  robot  visco-­‐ elastic  covering,  the  use  of  spherical  and  compliant  joints,  and  distributed  parallel  actuation   mechanisms  to  lower  the  effective  inertia  of  the  robot  near  the  end  effector.     Perceived  safety  describes  the  user’s  perception  of  the  level  of  danger  when  interacting  with   a   robot,   and   the   user’s   level   of   comfort   during   the   interaction.   Achieving   a   positive   perception  of  safety  is  a  key  requirement  if  robots  are  to  be  accepted  as  partners  and  co-­‐ workers   in   human   environments.   Perceived   safety   and   user   comfort   have   rarely   been   measured  directly.  Instead,  indirect  measures  have  been  used—  the  measurement  of  the   affective   state   of   the   user   through   the   use   of   physiological   sensors,   questionnaires,   and   others.  That  is,  instead  of  asking  subjects  to  evaluate  the  robot,  researchers  frequently  use   affective  state  estimation  or  questionnaires  asking  how  the  subject  feels  in  order  to  measure   the  perceived  safety  and  comfort  level  indirectly.     Questionnaires  can  be  used  to  compare  different  configurations  of  a  robot  (Figure  11).  The   results  may  then  help  the  developers  to  choose  one  option  over  the  other.  In  the  future,  this   set   of   questionnaires   could   be   extended   to   also   include   the   believability   of   a   robot,   the  
  • 20. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.20  /  56     enjoyment  of  interacting  with  it,  and  the  robot’s  social  presence.  However,  the  perceptions   of  humans  are  not  stable.  The  more  humans  get  used  to  the  presence  of  robots,  the  more   their  knowledge  and  expectations  might  change.  The  questionnaires  can  therefore  only  offer   a  snapshot  and  it  is  likely  that  if  the  experiment  would  be  repeated  in  twenty  years,  it  would   yield  different  results.  [17]       Figure  11.  An  example  of  a  measurement  questionnaire  about  robot’s  factors        
  • 21. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.21  /  56     Below  the  results  of  a  questionnaire  on  robot  acceptance  are  extracted.     An   interesting   observation   they   noticed   is   a   tendency   towards   considering   a   display   of   a   facial  expression  as  relatively  unimportant.  More  physical  factors  such  as  size  and  degree  of   anthropomorphic   design   seem   to   be   regarded   as   neither   extremely   important   nor   unimportant.   From   another   part   of   the   questionnaire   the   following   properties   seem   to   have   some   inherent  importance:   • The  robot  should  be  a  multipurpose  tool;   • Easy  to  instruct;   • It  has  to  behave  correctly;   • It  should  be  safe  to  use  and  induce  confidence;   • Properties  seem  to  be  of  lesser  importance:   • Having  a  personality;   • Being  humanlike;   • Engage  in  social  contact;     Among   the   more   indecisive   properties   as:   "Autonomy",   "Intelligence",   and   "Appropriate   Size"  there  are  no  indications  of  a  general  trend.     It  is  important  to  note  that  a  low  score  does  not  disqualify  a  certain  property  as  unimportant,   but  rather  that  the  ones  with  a  general  higher  score  are  judged  as  being  more  important  for   instigating  a  positive  feeling  towards  the  robot.  In  the  next  part  of  the  questionnaire  they   turned  the  issue  around,  asking  for  the  properties  that  were  most  important  in  creating  a   negative  feeling  towards  the  robot.  The  properties  chosen  were  in  this  part  also  chosen  for   their  possible  negative  connotation.     The  observable  tendencies  are  in  this  case  fewer  and  less  obvious.  The  clearest  triggers  of   negative  feelings  resulted  to  the  following  properties:   • It  does  not  understand  the  user;   • It  is  difficult  to  understand  its  actions;   • It  is  complicated  to  use;   Less  clear,  but  still  given  high  grades  by  several  subjects  were  the  following  factors:   • The  robot  stops  unexpectedly;   • The  robot  looks  heavy;   • The  robot  often  asks  the  user  how  to  proceed;     More  astonishing  might  be  that  a  machine-­‐like  appearance  of  the  robot  seems  to  generate   more  positive  feelings  than  a  human-­‐like  or  an  animal-­‐like.   Overall  most  of  the  factors  in  this  part  of  the  questionnaire  were  indicated  to  be  important   factors,   and   there   was   no   real   visible   trend   to   judge   any   one   property   as   more   or   less   important.  The  interpretation  of  this  is  that  most  of  these  properties  will  give  the  user  a   negative  feeling  towards  the  robot,  unless  they  are  tended  to  in  the  design.  [18]      
  • 22. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.22  /  56     Research  conclusions       Human-­‐robot  interaction  (HRI)  is  a  fairly  new  branch  of  HCI  (human-­‐computer  interaction)   and   has   gained   a   lot   of   attention   recently.   Concerning   a   mobile   service   robot,   additional   aspects  with  respect  to  user  acceptance  and  their  expectations  have  to  be  considered.  So,   what  are  people’s  views  on  the  role  of  an  intelligent  service  robot  in  their  home?    Different   studies   have   been   conducted   to   investigate   people’s   attitudes   towards   domestic   robots.   Syrdal  [19]  carried  out  a  survey  in  order  to  examine  adults’  attitudes  towards  an  intelligent   service  robot.  Participants  were  21-­‐60  years  old,  while  most  of  them  were  in  the  age  of  21-­‐ 30.   Results   show   that   most   of   the   participants   were   positive   towards   the   idea   of   an   intelligent  service  robot  and  view  it  as  a  domestic  machine  or  smart  intelligent  equipment   that  can  be  ‘controlled’,  but  is  intelligent  enough  to  perform  typical  household  tasks.  On  the   other  hand,  Scopelliti  [20]  investigated  people’s  representation  of  domestic  robots  across   three  different  generations  and  found  that  while  young  people  tend  to  have  positive  feelings   towards  domestic  robots,  elderly  people  were  more  frightened  of  the  prospect  of  a  robot  in   the   home.   Studies   within   the   European   project   COGNIRON   assessed   people’s   attitudes   towards   robots   via   questionnaires   following   live   human-­‐robot   interaction   trials   [21].   Responses   from   28   adults   (the   majority   in   the   age   range   26-­‐45)   indicated   that   a   large   proportion  of  participants  were  in  favor  of  a  robot  companion,  but  would  prefer  it  to  have  a   role  of  an  assistant  (79%),  machine/appliance  (71%)  or  servant  (46%).  Few  wanted  a  robot   companion  to  be  a  ‘friend’.  The  majority  of  the  participants  wanted  the  robot  to  be  able  to   do   household   tasks.   Also,   participants   preferred   a   robot   that   is   predictable,   controllable,   considerate   and   polite.   Humanlike   communication   was   desired   for   a   robot   companion,   however,   human-­‐like   behavior   and   appearance   were   less   important.   These   three   studies,   conducted   in   different   European   countries,   agreed   with   respect   to   the   desired   role   of   a   service  robot  in  the  home:  an  assistant  able  to  carry  out  useful  tasks,  and  not  necessarily  a   ‘friend’  with  human-­‐like  appearance.  These  considerations  led  to  the  definition  of  a  robot   companion  which  must  a)  be  able  to  perform  a  range  of  useful  tasks  or  functions,  and  b)   carry  out  these  tasks  or  functions  in  a  manner  that  is  socially  acceptable  and  comfortable  for   people   it   shares   the   environment   with   and/or   it   interacts   with.   [19]   This   creates   the   following   challenge   for   the   development   of   such   a   robot:   we   have   to   bridge   the   gap   between   functionality,   which   goes   along   with   hard   technological   properties   of   e.g.   an   industrial  robot,  and  social  acceptance,  which  goes  along  with  the  comfortable  design  of   e.g.  an  electronic  pet.  [16]    
  • 23. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.23  /  56     The  aim       The  aim  of  this  thesis  is  to  investigate  the  user’s  familiarity  of  different  design  alternatives   for  service  robots  according  to  the  “uncanny  valley”  hypothesis.   For   this   purpose   a   number   of   approximately   seven   factor   levels   of   the   robot   appearance   varying   in   human-­‐likeness   were   designed.   Also   a   questionnaire   comparing   the   different   alternatives  was  created  for  the  test  subjects.  An  empirical  study  design  was  developed  and   a  pretest  was  carried  out.  Pretest  results  of  the  influences  of  the  familiarity  and  likeness  will   be   given   in   a   graph.   Then   a   comparison   between   the   resulted   graph   and   the   one   of   the   uncanny  valley  completes  the  thesis.     Design  concept       The   concept   of   the   design   approach   in   this   study   is   based   on   the   idea   to   cover   as   many   significant   points   of   the   uncanny   valley   graph   as   possible   (Figure   12.).   For   each   point   a   different  appearance  alternative  is  designed.  Starting  with  an  industrial  machine-­‐like  robot,   each  gets  more  human-­‐like  than  the  previous.  This  is  the  expected  ranking  order  given  to   them  by  the  authors,  according  to  the  “uncanny  valley”  hypothesis.           Research  questions   Related   to   the   above   issues,   the   present   study   addressed   the   following   main   research   questions:     What  is  the  importance  of  robot  appearance  for  less  human-­‐looking  robots?   Do  people  prefer  more  human-­‐like  appearance  in  robots  that  they  interact  with?   Does  gender  factor  in  both  sides  –  robots’  and  test  subjects’,  influences  the  judgments  of  the   participants?     How  many  design  alternatives  are  needed  in  order  to  get  more  complete  results?      
  • 24. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.24  /  56       Figure  12.  A  simplified  graph  of  the  uncanny  valley,  on  which  the  concept  of  current  study  is  based     Design  requirements  and  limitations   According  to  the  concept  above,  the  development  of  robot  appearance  should  follow  a  line,   instead  of  a  curve,  excluding  the  hypothetic  valley  of  the  graph.  With  this  approach  in  each   design  should  be  added  a  human-­‐like  feature,  enough  to  make  it  distinguishable  as  a  more   human-­‐like  than  the  previous  one  (Figure  13.).   The  graph  starts  with  a  robot  which  is  relatively  machine-­‐like  in  appearance.  It  will  have  no   overtly  human-­‐like  features.  Then  the  development  will  go  through  a  humanoid  -­‐  a  robot   which  is  not  realistically  human-­‐like  in  appearance  and  is  readily  perceived  as  a  robot  by   human  users.  However,  it  may  possess  some  human-­‐like  features,  which  are  usually  stylized,   simplified  or  cartoon-­‐like  versions  of  the  human  equivalents,  including  some  or  all  of  the   following:   a   head,   facial   features,   eyes,   ears,   eyebrows,   arms,   hands,   legs.   The   important   principle  that  should  be  kept  in  the  design  line  is  that  any  feature,  if  possessed  in  a  robot   appearance,  should  be  more  human-­‐like  in  the  next  robot.  Both  may  also  have  wheels  for   locomotion  or  use  legs  for  walking.  The  graph  should  end  with  an  android  -­‐  a  robot  which   exhibits  appearance  which  is  as  close  to  a  real  human  appearance  as  technically  possible.     The  working  definition  of  robot  appearance  for  humanoid  robots  used  in  this  study  is  based   on  the  definitions  for  animated  agents’  appearances  adopted  by  Gong  and  Nass  [22],  and  for   android  robots  from  Mac-­‐Dorman  and  Ishiguro  [15].    
  • 25. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.25  /  56       Figure  13.  Line  of  design  alternatives  development       For   a   better   user’s   recognition   each   design   alternative   should   be   shown   in   identical   environment  performing  the  same  tasks.  In  this  way  the  robot’s  purpose  will  be  clear  for  the   participants,  so  they  won’t  be  confused  by  the  different  appearances.   The   styling   of   the   sketches   and   the   models   respectively,   incl.   colors,   shapes,   materials,   should   be   also   identical   in   the   developed   alternatives,   because   if   not,   there   might   be   unwanted  impact  on  the  users’  impression  about  them.  For  example,  if  a  box  shape  is  used   for   a   head   in   the   machine-­‐like   alternative,   then   a   box   shaped   head   should   be   used   in   a   similar  way  for  the  more  human-­‐like  alternative.     Design  project   Below  are  the  initial  sketches  for  six  of  the  alternatives.  The  performing  task  for  each  one  is   fetching  and  carrying  a  glass  and  a  bottle  of  water.     The  seventh  level  of  human  likeness  is  actually  a  real  human.  Note  that  for  humanoid  and   android   parts   of   the   graph   authors   have   developed   a   second   level   factor   alternatives   for   each   robot   –   a   male   and   a   female   gender.     This   altogether   makes   eleven   alternative   visualizations.    
  • 26. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.26  /  56         Figure  14.  Initial  sketches  for  design  alternative  Nr.1     Alternative  #1   The   main   concept   here   is   to   use   basic   shapes   (sphere,   cylinder)   and   simple   cuts   for   creating   a   robot  which  is  closer  to  an  object  executing  the  task   –  carrying  a  bottle.  What  is  crucial  here  is  that  this   object  should  look  like  it  is  moving  itself,  and  not  just   standing  (as  the  participants  are  going  to  see  only  a   picture,  not  an  animation).  That  is  why  appearance   #1   should   definitely   have   wheels,   or   chains,   and   if   wheels  –  then  preferably  more  than  two,  for  more   confidence  in  its  stability.          
  • 27. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.27  /  56         Figure  15.  Initial  sketches  for  design  alternative  Nr.2   Alternative  #2   Here  the  concept  resembles  the  one  of  Care-­‐o-­‐bot  –   the   two-­‐side   concept   (working   and   serving   side).   That  means  assuming  there  is  at  least  one  hand  or   some  other  part  for  fetching  objects.  Still,  it  is  not  an   android   or   a   humanoid,   but   its   silhouette   could   remind  of  a  short  waiter.  A  display  for  serving  side  is   preferable,   as   well.   Movement   could   be   two-­‐   or   three-­‐wheels,   but   again   because   of   stability   and   easier   recognition,   a   wheel   system   in   the   back   can   play  the  role  of  stabilizer.      
  • 28. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.28  /  56         Figure  16.  Initial  sketches  for  design  alternative  Nr.3     Alternative  #3   As   robots   get   more   and   more   human-­‐like,   in   alternative   #3   we   can   already   nave   some   more   human   features   like   head   and   two   arms,   for   example.   In   this   case   the   head   is   rather   a   helmet,   and  a  big  mask  instead  of  a  display  or  face.  Two  arms   or  a  tray,  attached  to  the  body  is  the  functional  part   which   the   patient   interacts   with.   Proportions   are   childish  –  bigger  head  and  smaller  body.    
  • 29. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.29  /  56       Figure  17.  Initial  sketches  for  design  alternative  Nr.4     Alternative  #4   The  concept  is  developing  smoothly  to  a  humanoid,   adding   legs,   which   may   be   separate,   or   just   imitating.  Here  we  can  already  have  a  neck,  attached   to   a   head   or   a   helmet.   Smaller   mask   and   more   detailed  arms,  as  well  as  shoes/feet  are  preferable.    
  • 30. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.30  /  56         Figure  18.  Initial  sketches  for  design  alternative  Nr.4,  part  2     Alternative  #4   The   concept   of   alternatives   from   #4   onwards   will  differ  in  genders,  as  well.  In  this  design  the   concept   changed   its   direction   to   an   animation   character’s  proportion  for  the  humanoids.  This   means   really   exaggerated   big   head   and   significantly   smaller   body   in   different   from   human  proportions.    
  • 31. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.31  /  56                 Figure  19.  Initial  sketches  for  design  alternative  Nr.5     Alternative  #5   Approaching  the  android  we  risk  falling  into  the  uncanny  valley.  That’s  why  we  want  to  make   this  falling  controlled,  by  switching  to  exaggerated,  but  rather  detailed,  human  features,  but   still   following   the   line   of   human   likeness.   For   example   –   making   the   eyes   and   the   mouth   extremely  big  is  acceptable,  but  placing  the  face  on  the  body  part  is  not  –  it  is  less  human  like   than  the  previous  alternatives.    
  • 32. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.32  /  56                                         Figure  20.  Initial  sketches  for  design  alternative  Nr.6       Alternative  #6   For  the  right  hill  of  the  uncanny  valley  graph  we  have   and   android   which   should   possess   already   human-­‐ like   detailed   features   like   real-­‐size   eyes   and   an   appropriate  nice  sight.  Nose  and  ears,  as  well  as  an   open  mouth  can  finish  the  face.  Their  arms  and  feet   can  be  supplied  with  fingers,  instead  of  just  clippers.   Longer  legs  and  knees  would  increase  the  impression   of  available  realistic  movement.      
  • 33. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.33  /  56     Pretest  design   Scenario   The   context   chosen   for   the   study   and   associated   HRI   trials   was   that   of   a   domestic   robot   bringing  a  glass  and  a  bottle  of  water  to  the  participant.     Participants  and  procedure   At  the  beginning  of  each  trial,  information  about  the  purpose  of  the  current  test  was  given   to  each  of  the  participants  and  detailed  instructions  for  participating  in  the  experiment.  A   supervisor  was  on  hand  to  answer  any  further  questions  and  to  repeat  the  instructions  if   necessary.     Twenty  participants  took  part  in  the  pretests.  There  were  mainly  students  (60%),  7  female   and   13   male.   The   average   age   was   26.   Everyone   could   communicate   in   English   with   12   growing  up  in  Germany,  and  8  growing  up  in  other  non-­‐English  speaking  countries  (4  in  India,   3  in  Spain,  1  in  Turkey).  The  majority  (80%)  said  they  had  some  information  about  any  kind   of  robots  –  incl.  rescue  robots,  service  robots,  industrial  building  robots,  but  only  30%  of   them  had  real  live  experience  with  robots.  The  participants  were  all  volunteers  and  none   received  remuneration.     The  pretests  took  place  in  a  university  laboratory.  The  test  subjects  were  sitting  in  front  of   24”  LCD  screen.  Only  one  subject  at  a  time  was  examined.  Each  of  them  completed  the  first   part   of   a   questionnaire   individually,   providing   basic   demographic   details   including   background,  gender,  and  age,  as  well  as  robot  information  or  interaction  experience  before   they   were   exposed   to   the   testing.   The   relevant   questions   from   the   questionnaire   are   provided   below   in   the   Questionnaire   section.   The   participants   were   then   shown   visualizations  of  eleven  robots.  Robots’  appearances  were  labeled  with  numbers  only  for  the   supervisor   in   order   to   avoid   any   possible   influences   on   the   subjects’   judgments.   Each   participant  was  given  a  different  sequence  of  the  robots’  pictures.  In  this  way  there  is  no   possibility  for  them  to  guess  our  predicted  initial  ranking  order  of  the  alternatives  as  they   were   created.   The   participants   had   to   fill   in   the   second   part   of   the   questionnaire   while   watching  the  pictures  in  order  to  mark  their  preferences  towards  each  robot  appearance.  A   sample  of  the  test  sheet  is  shown  in  the  Questionnaire  section,  as  well.  Each  of  the  two   20cm-­‐long  horizontal  lines  represents  a  100%  scale  for  both  factors  –  Human  likeness  and   Familiarity.  For  the  upper  one  (Human  likeness)  the  extremities  are:  0%  for  Machine-­‐like,   and   100%   for   Human   like   appearance   of   the   visualized   robot.   For   the   second   line,   respectively:  0%  for  Very  strange,  and  100%  for  Very  familiar.  Each  participant  had  to  mark   (using  a  vertical  line  or  a  cross)  on  each  line  their  evaluation.     Finally,   the   participants   were   asked   whether   they   had   any   difficulty   completing   the   questionnaire  or  understanding  what  they  were  supposed  to  do  in  the  experiment;  and  what   their  impression  was  of  the  11  robot  designs.     All   evaluations   from   the   second   part   of   the   questionnaire   were   then   measured   in   centimeters  and  the  data  was  transferred  into  a  table  (Table  1.).      
  • 34. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.34  /  56     Visualizations   Visualizations  were  designed  and  rendered  in  CAD  software  –  3D  Studio  Max  Design  2012.   At  first,  all  the  robots  should  be  visualized  in  identical  environment  and  doing  the  same  task   –  in  this  case  they  carry  a  tray  with  a  bottle  and  a  glass  of  water.  According  to  some  findings   of   Walters   et   al.   [23],   humans   strongly   did   not   like   a   direct   frontal   approach   by   a   robot,   especially   while   sitting   (even   at   a   table)   or   while   standing   with   their   back   to   a   wall.   An   approach  from  the  front  left  or  front  right  was  preferred.  In  this  case,  all  the  visualizations   are  taken  with  the  robots  “approaching”  from  the  front  right  side  of  the  screen.       Design  alternative  #1  (Figure  21.)  has  been  chosen  to  resemble  a  simplistic  association  with   the  famous  R2D2  robot.  This  is  because  it  is  familiar  enough,  and  still  does  not  possess  any   human  features,  but  rather  is  closer  to  an  object  from  a  living  environment  (e.g.  a  small  table   or  a  mobile  mini-­‐bar).    The  function  concept  behind  is  that  the  hemisphere  on  the  top  can   open  from  the  needed  side  (either  from  user’s  side,  or  working  side  –  where  bottle  charging   happens).  The  robot  moves  by  3  spherical  wheels,  symmetrically  placed  on  its  bottom.   Design  alternative  #2  (Figure  22.)  won  among  the  rest  of  the  sketch  concepts  because  its  lack   of  a  hand  makes  the  curve  of  transition  between  alternatives  smoother.  Still  its  silhouette   and   movement   signs   remind   more   of   an   object   than   a   human.   It   possesses   a   display   for   interaction   with   the   user.   The   wheels   are   also   three,   but   cylindrical   and   asymmetrically   places,  two  of  which  in  the  back.  This  was  imposed  by  the  asymmetrical  shape,  which  makes   the  silhouette  to  have  a  direction  facing  the  user  or  the  working  plot.  In  non-­‐active  state  the   display  can  be  closed.   Design   alternative   #3   (Figure   23.)   is   designed   to   have   rather   baby   proportions   –   very   big   head   and   a   small   body,   short   arms   and   posing   like   squatting   or   crawling.   That’s   why   the   wheels  here  are  already  four  (association  with  a  baby  crawling).  Its  spherical  helmet  play  the   role  of  a  head,  together  with  a  big  mask,  looks  more  like  a  space  suit.  The  tray  in  this  case  is   attached  to  two  short  “arms”  which  could  be  height-­‐adjusted.   Design   alternatives   #4   (Figures   24.25.)   are   already   in   two   versions   –   female   and   male.   Cartoons  –  a  bigger  head  and  tiny  body,  inspire  their  proportions.  The  head  is  still  spherical   in  shape,  keeping  the  style  from  the  previous  alternative  #3.  There  is  a  darker  helmet,  which   can  be  considered  as  hair.  Here  the  mask  is  smaller,  resembling  sunglasses,  slightly  different   depending  on  the  gender.  Mouth  appears  on  the  face  to  make  it  more  character  look-­‐alike.   More  humanoid  features  come  up  –  arms  are  proportional  to  the  body  and  are  separate   from   the   tray.   The   hands   are   like   clippers.   The   female   has   a   “dress”,   while   the   male   is   “dressed”  in  trousers.  Their  moving  equipment  consists  of  small  “shoes”,  and  it  is  supposed   to  slide  on  the  ground,  instead  of  rolling  wheels.   Design   alternatives   #5   (Figures   26.27.)   go   closer   to   the   predicted   uncanny   valley   of   the   graph.   That’s   why   the   concept   of   exaggerating   human   features   is   applied   here.   It   is   expressed   most   noticeable   in   face   features   like   bigger   eyes   and   wider   mouth,   small   and   animal-­‐type  nose.  Gender  signs  are  obvious  on  the  upper  bodies.  Limbs  are  turned  up  side   down   –   wider   in   the   lower   parts.   But   the   extremities   here   are   more   detailed   than   the   previous  alternative  –  they  have  knees  and  elbows,  as  they  are  even  moving  independently   (not  sliding  together).    
  • 35. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.35  /  56     Design  alternatives  #6  (Figures  28.29.)  are  going  up  to  the  android  on  the  predicted  graph  in   this  study.  That’s  why  face  features  are  much  closer  to  the  real  human  eyes,  mouth,  nose,   and   even   cheeks,   as   well   as   the   head   shape,   which   is   already   looking   like   human   skull.   Eyebrows  and  ears  are  added.  The  extremities  are  in  proper  human  proportions.  Hands  have   fingers  and  feet  are  more  human-­‐like.     Design  alternatives  #7  (Figures  30.31.)  are  supposed  to  be  the  real  humans.  They  are  re-­‐ defined   three-­‐dimensional   models,   adapted   to   the   style   of   the   previous   alternatives.   The   most   important   here   was   their   sight   –   it   should   be   nice   and   friendly,   not   frightening   the   users.  This  task  is  very  difficult,  because  it  is  still  not  a  real  human,  and  it  is  obvious  for  the   test  subject,  who  is  a  real  human.  Both  female  and  male  models  present  nice  looking  waiters   or  companion  staff.  Their  clothes  are  in  same  colors  like  the  robots,  so  test  subject  would   not  be  confused  of  too  different  factor  levels  (like  different  colors).                 Figure  21.  3-­‐dimensional  model  for  design  alternative  Nr.1  
  • 36. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.36  /  56         Figure  22.  3-­‐dimensional  model  for  design  alternative  Nr.2  
  • 37. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.37  /  56       Figure  23.  3-­‐dimensional  model  for  design  alternative  Nr.3    
  • 38. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.38  /  56       Figure  24.  3-­‐dimensional  model  for  design  alternative  Nr.4  –  female  version    
  • 39. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.39  /  56       Figure  25.  3-­‐dimensional  model  for  design  alternative  Nr.4  –  male  version  
  • 40. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.40  /  56       Figure  26.  3-­‐dimensional  model  for  design  alternative  Nr.5  –  female  version    
  • 41. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.41  /  56       Figure  27.  3-­‐dimensional  model  for  design  alternative  Nr.5  –  male  version    
  • 42. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.42  /  56         Figure  28.  3-­‐dimensional  model  for  design  alternative  Nr.6  –  female  version    
  • 43. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.43  /  56         Figure  29.  3-­‐dimensional  model  for  design  alternative  Nr.6  –male  version    
  • 44. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.44  /  56       Figure  30.  3-­‐dimensional  model  for  the  real  human  –  female  version      
  • 45. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.45  /  56       Figure  31.  3-­‐dimensional  model  for  the  real  human  –male  version      
  • 46. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.46  /  56     Questionnaire       no.  of  participant:             date:             time:     First,  we  ask  you  for  some  information  about  yourself.  Thank  you!     1.  How  old  are  you?   ______  years  old     2.  What  is  your  gender?   □  Male        □  Female     3.  What  is  your  highest  level  of  education?   □  No  graduation  □  secondary  modern  school        □  PhD     □  Business  school  □  On-­‐the-­‐job  training            □  A-­‐level   □  Studies          □  secondary  school  level  I  certificate  □  Other:  _________     4.  Where  do  you  work  /  you  have  worked?   □  Craft                □  Social-­‐/Humanities  sciences   □  Mercantile  /Administration  □  technical/Natural  science   □  Housewife/-­‐husband        □  Other:  _______________________     5.  Have  you  ever  read,  heard  or  seen  anything  about  robots     (e.g.  in  newspapers,    books,  television,  movies  or  from  friends)?   □  No   □  Yes,  about  following  robots    __________________________________   __________________________________   __________________________________   __________________________________     6.  Did  you  have  already  gained  some  experience  by  using  robots?   □  No      □  Yes,  with        __________________________________   __________________________________   __________________________________  
  • 47. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.47  /  56        
  • 48. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.48  /  56     Pretest  results  and  analyses   Below  all  the  data  collected  from  the  pretests  is  extracted  (Table  1.).     First,  the  average  ratings  for  Human  likeness  and  Familiarity  were  measured  (Figure  32.).   The   standard   deviations   are   also   shown   there.   In   our   case   there   are   some   big   values   for   some  of  the  alternatives  (mostly  for  extremities  –  the  machine  like  and  the  real  humans).   Regarding  the  low  numbers  of  test  subjects  in  the  pretests,  we’d  rather  conclude  that  the   standard   deviation   shows   us   that   people   have   very   different   opinions   about   the   robots’   appearance.  These  deviations  could  be  in  a  different  range  after  the  real  tests  later.     Figure  32.  Human  likeness  and  Familiarity  curves  with  Standard  deviations  measured  on  11-­‐point  scale       An  important  data  about  the  differentiations  in  robots’  genders  could  be  seen  in  the  graph   above.   It   shows   very   close   results   between   the   female   and   male   forms   in   each   of   the   alternatives.   This   means   that   we   don’t   actually   need   these   differentiations   between   the   genders  and  we  could  reduce  the  number  of  alternatives  to  7,  instead  of  11.    This  reduces   the  time  needed  for  running  the  real  tests  as  well.   Then  comes  the  question:  Which  gender  should  we  take  away?   The  answer  comes  from  the  two  graphs  below  –  Figure  33  and  Figure  34.      
  • 49. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.49  /  56       Table  1.  Data  results  from  the  pretests  
  • 50. RHI:  Investigation  of  the  uncanny  valley     Alessya  Ivanova     p.50  /  56       Figure  33.  The  uncanny  valley  graphs  for  both  genders  of  the  robots       Figure  34.  Human  likeness  curves  for  both  participants’  genders     The  Uncanny  valley  for  robot’s  genders