Crowdsourcing Activism discusses using bots to recruit citizens to participate in activism. It describes how activists currently use social media to share their vision and recruit support but it takes a lot of time. Governments and large organizations have been using bots to influence discussions. The paper proposes Botivist, which uses different bot strategies like direct requests, solidarity, gain, and loss to recruit citizens in a cause. A user study found direct strategies generated the most responses and volunteers while solidarity strategies got the most retweets and favorites. Different citizen types engaged more with certain strategies. The research shows bots can recruit citizens for collective action and different audiences react to strategies differently from humans or bots.
2. Activismo y Redes Sociales
Redes sociales ya son usadas por los activistas para compartir su
visión y reclutar cuidados para su causa. Pero activistas tienen
que invertir mucho tiempo en estas tareas.
2
3. 3
Bots y Redes Sociales
Gobiernos y grandes organizaciones con mucha experiencia han
estado usando los bots para callar discusiones, persuadir y
cambiar enfoque.
4. Problemas
! Usar redes sociales estratégicamente es complicado: requieres
personal con mucha experiencia.
! Reclutar gente para una causa social es dificil:
– toma tiempo
– es tedioso
– poco productivo.
4
6. Botivist
6
Ciudadanos son reclutados para
contribuir a una causa social.
Causa
Social
Estrategia A Estrategia B
Estrategia DEstrategia C
Botivist
7. Estrategias de los Bots
7
Estrategia A: Directo
Estrategia B: Solidaridad
Estrategia C: Ganancia
Colaboramos para combatir la
corrupción? Cómo reducimos la
corrupción en nuestras calles?
Colaboramos para combatir la corrupción?
Cómo reducimos la corrupción en nuestras
calles? Hazlo por ti, por mi, por México!
Colaboramos para combatir la corrupción?
Cómo reducimos la corrupción en nuestras
calles? Juntos podemos mejorar México!
Estrategia D: Perdida
Colaboramos para combatir la corrupción? Cómo
reducimos la corrupción en nuestras calles? Sin
colaboración, el futuro de México es negro!
Usuario : CherryRex
Contraseña: Pi3.14080878
8. Botivista: Funcionalidad
8
1. Identifica posibles voluntarios de
Tweets que usan ciertas palabras claves
2. Bots mandan tweets a ciudadanos pidiendo
acción inicial usando cada estrategia.
Estrategia A Estrategia B Estrategia C
3. Bots ensamblan activismo con ciudadanos reclutados, guiándolos y pidiendo
micro-tareas para la causa.
#Ayotzinapa6Meses este
gobierno corrupto es el
responsable ya renuncien!
Descubren mentiras del
corrupto de Osorio Chong!
Estrategia A Estrategia B Estrategia C
11. Metodología:
Estudiando Participación Ciudadana
11
! Cada bot interaactuó con grupos diferentes de ciudadanos.
! Estudiamos tipo de participación ciudadana creada por cada estrategia.
! Analizamos:
– numero de respuestas o “replies”
– numero de retweets, number de favorites
– numero de personas participando
!
Corrimos una prueba anova, asi como comparaciones directas por pares para
ver si había diferencia significativa entre estrategias.
12. Resultados
12
*Bots que son directos generaron:
! Mayor número de respuestas de los
ciudadanos.
! Mayor numero de voluntarios únicos.
! Mayor numero de interacciones entre
ciudadanos.
Strategy
*Bots que muestran perdidas generaron:
•Mayor numero de interacciones entre
ciudadanos
*Bots con Solidaridad generaron:
! Mayor numero de retweets y favoritos
al contenido del bot por parte de
ciudadanos.
Numero de Respuestas
Ciudadanas por Tweet del Bot
Numero deVoluntarios
Numero de Interacciones
Ciudadanas por Tweet del Bot
Numero de Interacciones
entre Ciudadanos
13. Metodología:
Estudiando Calidad de Contribuciones
13
! Estudiamos calidad de participación ciudadana creada por cada estrategia.
! Analizamos:
– La contribución del ciudadano es relevante a lo que pido el bot?
!
Reclutamos 3 crowd workers para categorizar cada respuesta de los
ciudadanos en si era relevante o no: contribuye una idea para compartir la
corrupción?
14. Results
14
! Overview per strategy of the percentage of on-topic audience
members.
Majority of contributions made by
recruited audiences were relevant for
their task.
* Almost 100% of the audience
members who engaged in Direct
Strategy made relevant contributions.
*Audience members engaging in Loss
Strategy made the most irrelevant
contributions.
15. Methodology:
Most Active Audience Members
15
! We discover common traits of the most active
audience members recruited by Botivist.
Identify highly active
audience members.
Characterize highly
Use Mean Shift to cluster active
audience
16. Results: Traits Most Active Audience Members.
16
The Negative Nationalists
• Most tweets used nationalistic terms and had negative sentiment.
• Most audience members recruited by Gain modality.
• “Negative” people recruited best with automatic agents who share
“hope” messages.
The Community Nationalists
• Most tweets were about social issues and nationalism.
• Most active of all.
• Majority were recruited with the Loss or Solidarity mode.
• Showing solidarity and what could be lost without participation was
effective to recruit nationalistic concerned citizens.
The Short Lived Activist
• Less than 3% of their tweets referenced political content or
social issues.
• Majority recruited by the direct modality.
• For Individuals with no political affiliation, direct messages most
effective to foster their participation.
18. Identificando Diferencias entre
Ciudadanos
18
1.Tomamos todos los
tweets personales de
personas que bots
tratamos de reclutar.
2. Usamos un Mann-Whitney
rank test. Para encontrar
palabras (hashtags, usuarios,
palabras) que son usados por
un grupo mas que otro.
3. Categorizamos
palabras encontradas
usando Topic Modeling
+ Crowdsourcing
4. Medimos que
tanto cada grupo
habla de cada
categoría
encontrada.
20. 20
Diferencias entre Ciudadanos que
Participan con Bots y los que no.
Ciudadanos que responden a
bots tienen a usar palabras
sobre activismo, noticias y
marketing!
Ciudadanos que NO
responden a bots tienen a usar
palabras sobre política y
noticias.
21. Botivist Research Takeaways
21
! Automated agents can be used to recruit citizens, and
incite collective efforts for an activists’ cause.
! Citizens react to strategies differently when coming from
humans or automated agents.
! Specialized citizens engage more with specific strategies,
e.g., solidarity. Citizens with general skills participate more
with more general direct strategies.
25. Traditional Media Content Distribution
25
! Information is filtered
through hierarchal
organizations before
reaching the audience.
The organizations focus
primarily on commerce.
Spreadable Media, H. Jenkins, et al.
"We the Media”, D. Gilmor
Gate keepers (e.g., Advertisers)
Organization (e.g., TV Channel)
Content (e.g., TV Shows)
Audience
26. Social Media Content Distribution
26
! Participants are peers and can change roles.
! Content is unfiltered before reaching the audience.
Community
Audience
Reporter Publisher
Reporter
Advertiser
Editor
Community
Spreadable Media, H. Jenkins, et al.
"We the Media”, Dan Gilmor
27. Problem
27
We lack understanding of the new relationships, tensions,
experiences emerging between audience & content
producers in social media.
*Pasquali F. et al.,“Emerging Topics in the Research on
Digital Audiences and Participation,”
29. Example: Problematic Design
29
Facebook gets sued!
Facebook designed in 2013 “organic” interactions with companies via side stories.
User A Company
User B receives
sponsored story
30. .
30
! I use social media to understand the experiences,
relationships, tensions, and interactions emerging from
content producers and their online audience.
Friendly-Intimate
Spaces
Adverse
Spaces
Controversial
Spaces
!
I use the understanding to design novel tools to
better engage with online audience.
*A Rhetoric Of Motives, Burke, ”
31. Main Research Findings
31
! Online audiences and content producers interact in a
gift economy focused on reciprocity and collaboration.
Multi-faceted data visualizations and online
autonomous agents are tools that can facilitate
reciprocity and collaborations between audiences and
content producers.
32. Impact
! Opens design space of systems for engaging
online audiences which focus on cooperation and
reciprocity over profit.
! Interactive systems focused on using the
intelligence of the audience.
32
Jenkins, H. "Interactive audiences? The collective intelligence of media fans.
Baym, N. et al., "Amateur experts International fan labour in Swedish independent music."
33. Talk Outline
! Problem
! Contribution
! Platforms for Engaging Online Audiences
! Visualizing & Engaging Online Audiences
! Engaging Online Audiences with Automated Agents
! Engaging Online Audiences Opportunistically
! Online Audiences — Future Work
33
34. Engaging Online Audiences
34
! I proposes two system designs to engage online
audiences:
(1) Authors understand in detail their audience
and use that knowledge to engage and
collaborate with them.
(1) Authors don’t know anything
about audience. Let bots to the
work!
! Visualizing Audiences ! Automated Agents
Person Visualizes+Understands! Use Knowledge to Engage!
Let the bots do
all the work !
Person understands
audience in detail.
35. Visualizing Online Audiences
Savage S., et al.,Visualizing Targeted Online Audiences,
COOP’14: Conference on the Design of Cooperative Systems.
36. Visualizing Online Audiences
36
• Long lists make it difficult to gauge
the traits of audience to motivate collaborations.
Need for:
– interfaces that facilitate
understanding one’s
audience to motivate
support, collaborations
and reciprocity.
Published: COOP’15
38. Design Proposals
38
! Human in the loop interfaces to target
audiences.
! Multifaceted data visualizations to help
creators target audiences for their
different collaborative tasks.
! Systems that let creators probe
different strategies to recruit and call
audiences to action.
39. Diversity Workflow
39
39
Interest Detection
People’s Online Profiles
User Modeling
Input
People’s Tweets
Interest 1: Music
Likes: Orange
Interests:
Bobby
#yaMecanse5 liberen el peje!
Interests
Visualization Engine
Interest: Pets Interest: Tech
User Modeling
Data Visualizations
40. 40
Savage S., et al.,Visualizing Targeted Online Audiences,
COOP’14: Conference on the Design of Cooperative Systems.
46. Hax Evaluation Methodology
46
! Between subjects study (N=15). Participants either
used Hax or Facebook’s traditional interface to
motivate audiences for a set of causes.
! Surveyed and interviewed participants on their
experiences, strategies adopted to complete the tasks,
benefits and drawbacks they saw, and a comparison
! between Hax/Facebook and other tools.
47. Hax Results
47
! Participants preferred Hax over list-based interfaces.
! Participants identified Hax facilitated new interactions
with audiences:
– Serendipitous Discoveries
– Facilitate Diffusion and Participation
– Audience Diversity
– AudienceVerification
48. Talk Outline
! Problem
! Contribution
! Platforms for Engaging Online Audiences
! Visualizing & Engaging Online Audiences
! Engaging Online Audiences with Automated Agents
! Engaging Online Audiences Opportunistically
! Online Audiences — Future Work
48
50. Botivist: Calling Online Audiences to Action
50
! Probes different strategies to recruit and initiate
collaborations with online audiences.
Solidarity
Gain
Loss
Direct
Could we collaborate to fight corruption?
One for all, and all for one!
Could we collaborate to fight corruption
to help improve our cities?
Could we collaborate to fight corruption? if
not the future of our cities will be grim.
Could we collaborate to fight corruption?
51. Methodology: Analyzing Audience Participation
51
! Between subject study on
Twitter to understand the type
of audience participation each
strategy generated.
! Analyzed:
– number of replies
– number of retweets, number of
favorites
– number of people participating.
! Ran an anova test, and pairwise
comparisons to see if there is
significant difference between
strategies.
52. Results
52
*Being Direct generated:
! the most replies from audiences;
! most unique number of volunteers;
! most number of interactions between
audience members.
! Overview of the number of audience
members and contributions which
each strategy triggered.
*
**
*
*
Strategy
*Showing Losses generated:
! high number of interactions among
audience members.
*Having Solidarity generated:
! Most number of retweets and favorites to
bot’s content.
53. Results
53
! Overview per strategy of the percentage of on-topic audience
members.
*Majority of contributions made by
recruited audiences were relevant for
their task.
* Almost 100% of the audience
members who engaged in Direct
Strategy made relevant contributions.
*Audience members engaging in Loss
Strategy made the most irrelevant
contributions.
54. Methodology:
Most Active Audience Members
54
! We discover common traits of the most active
audience members recruited by Botivist.
Identify highly active
audience members.
Characterize highly
Use Mean Shift to cluster active
audience
55. Results: Traits Most Active Audience Members.
55
The Negative Nationalists
• Most tweets used nationalistic terms and had negative sentiment.
• Most audience members recruited by Gain modality.
• “Negative” people recruited best with automatic agents who share
“hope” messages.
The Community Nationalists
• Most tweets were about social issues and nationalism.
• Most active of all.
• Majority were recruited with the Loss or Solidarity mode.
• Showing solidarity and what could be lost without participation was
effective to recruit nationalistic concerned citizens.
The Short Lived Activist
• Less than 3% of their tweets referenced political content or
social issues.
• Majority recruited by the direct modality.
• For Individuals with no political affiliation, direct messages most
effective to foster their participation.
56. Engaging Online Audiences: Botivist
Research Takeaways
56
! Automated agents can be used to recruit online audiences,
and incite collective efforts for an author’s cause.
! Audiences react to strategies (gifts) differently when
coming from humans or automated agents.
! Specialized audiences engage more with specific strategies,
e.g., solidarity. More general audience participate more
with more general direct strategies.
57. Engaging Online Audiences
Research Takeaways
57
! Multifaceted data visualizations help authors identify
strategies to motivate collaborations with their audience.
! Autonomous Agents help authors to probe strategies to
motivate and start collaborations with their audience.
58. Impact
! Opens design space of systems for engaging
online audiences which focus on cooperation and
reciprocity over profit.
! Interactive systems focused on using the
intelligence of the audience.
58
Jenkins, H. "Interactive audiences? The collective intelligence of media fans.
Baym, N. et al., "Amateur experts International fan labour in Swedish independent music."
59. Talk Outline
! Problem
! Contribution
! Platforms for Engaging Online Audiences
! Visualizing & Engaging Online Audiences
! Engaging Online Audiences with Automated Agents
! Engaging Online Audiences Opportunistically
! Online Audiences — Future Work
59
60. Savage S., et al., I'm Feeling LoCo: A Location Based Context Aware Recommendation
System, Lecture Notes in Geoinformation and Cartography. Springer.
Engaging Online Audiences
Opportunistically
61. Goals
A tool that facilitates volunteering & contributing
opportunistically:
• Understands users’ lifestyle and preferences.
• Understands users’ current context (activity)
• Match tasks to available and interested users.
61
68. 68
Infrastructure to Study Online Collective Action Ecosystem
• Scientific Framework to compare a collective effort’s results to how the effort
was organized.
• Visualizations to compare collective efforts across different axis.
• Human-in-the-loop interfaces to correct, and incorporate external knowledge.
• Allow scientific community to develop collective action principles.
69. 69
Theory of Design for Collective Action Systems
• Study how interface designs (data visualizations, wearables) affect
collaborations and computer based collective action.
• Present design principles for collaborative and collective action systems.
70. 70
The Future Generation of Computer Collective Action
• Study platforms that use big data to create end-to-end
computer based collective action systems. Impacts:
Smart Cities
Healthcare
Education
Government and Non Profits
Art
75. Selected Publications
75
Understanding Online Audiences
• Participatory Militias:An Analysis of an Armed Movement's Online Audience, CSCW’15
• Tag Me Maybe: Perceptions of Public Targeted Sharing on Facebook, HyperText’15
• The "Courage For” Facebook Pages:Advocacy Citizen Journalism in the Wild.
C+J:Computation+Journalism Symposium 2014
• Online Social Persona Management, U.S. Patent Disclosure, filed, 2013
System for Engaging Audiences
• Botivism: Using Online Bots to CallVolunteers to Action, CSCW’16 (under review)
• Visualizing Targeted Audiences, COOP’14
• I’m Feeling LoCo,A Location Based Context Aware Recommendation System,
Symposium of Location Based Services’11
• Socially and Contextually Appropriate Recommendation Systems,
U.S. Patent Disclosure, filed 2014.
• Search on the Cloud File System, PDCS’11: Parallel and Distributed Computing and Systems Conference
• Traversing Data Using Data Relationships, U.S. Patent Disclosure,
filed 2012, published 2014.
• CrowdsourcingVolunteers. Celebration of Women in Computing in Southern California 2014.
• An Intelligent Environment for User Friendly Music Mixing,IE’12: International Conference on Intelligent Environments
• Mmmmm:A multi-modal mobile music mixer,NIME’10: Conference on New Interfaces for
• Musical Expression
77. 77
Evaluation
Savage S., et al., I'm Feeling LoCo: A Location Based Context Aware
Recommendation System, Lecture Notes in Geoinformation and Cartography. Springer.
Launched our tool to the public. Study opportunistic participations
that our tool facilitates. Interviews and surveys to understand
subjective perceptions.
78. Understanding Online Audiences
Research Takeaways
78
! Friendly Environments
-Authors & Audiences collaborate to distribute surprising content and overcome algorithms.
-Authors tag as a strategy to harvest supportive audiences.
• Controversial Environments
-Authors use the strategy of adapting their self-presentation for their audience to encourage
more participation.
• Adverse Environments
-Authors collaborate with their audience to produce news reports and even offline
collective efforts.
-Authors have strategies to engage their audience: show solidarity by explaining how to keep
safe, use offline events to recruit newcomers.
80. Tag Me Maybe : Integrating
the Audience into Content
80Published: Hypertext’15
Tags
81. Method
81
Study
Participants are
interviewed
and surveyed
Lessons
Learned
Results
Responses are
Categorized and
Quantified
• Interview Questions focused on people’s perspectives on publicly
tagging friends in Facebook posts from the view of:
– Content Producers (Taggers)
– Audiences Tagged (Taggees)
– Passive Audiences (Viewers)
• Conducted qualitative coding to categorize interview responses.
82. Demographics Participants
82
Total number of participants 270
Participants recruited from FB 32
Participants recruited outdoors 88
Participants recruited from Amazon
Mechanical Turk
150
Sex Demographics 43% Female, 57% Male
Age Demographics 18-68 years old, median age of 22
86. Method
86
1. Detect different ways content
producers present themselves
in a controversial community.
2. Compare Self-Presentation
with content popularity based
on number of comments.
3. Interview participants from
each cluster to further
understand the behavior.
Input
Topic Modeling
People’s PostsPeople’s Profiles
User Modeling
Clustering
}
Different types of
self presentations
87. LiveJournal Data
87
! Collected LiveJournal (LJ) Posts from ONTDP and Profiles from
all bloggers to ONTDP from March 30th 2012 to July 11th 2012.
Authors 296
Commenters 1,972
Interviewees 12
LJ posts 1,200
Comments 30,934
Profile Tags 9,812
Post Tags 1,622
88. Results
88
Content producer’s whose self presentations were tailored
to the audience received the most attention and
participation from the audience.
People who got most
comments from their
audience.
89. Understanding Online Audiences:
Controversial Environments
89
! Content Producers use the strategy of adapting
their self-presentation to audience’s interests to
obtain more participation.
90. Participatory Militias
90
Armed civilian forces have
successfully fought back against
organized crime.
These groups have been active on
social media, particularly on
Facebook pages titled “Courage for
X Region” to inform and recruit
citizens to resist the criminals.
Published: CSCW’15
91. Data “Courage For” Facebook
Posts: 25,878
Fans: 488,029
Comments:108,967
Post Likes: 1,481,008
Reshares: 364,660
92. Goal
Use the “Courage For” pages as a lens to understand
1) the type of online content shared by content
creators in adverse scenarios and how the online
audience engages with content creators;
2) characteristics of the most active audience
members.
93. Methodology Topic Understanding
93
! We used a grounded theory approach to identify the
main topics inVXM’s posts.
1. Extracted topics from set of 700 randomly selected posts.
2. Used oDesk to hire three Spanish-speaking, college educated
people to categorize theVXM posts.
3. We used a majority rule to determine the topic each post.
Topic 4
Topic 2
94. What do Content Producers
Share?
94
! Primarily News
Reports, but also
content to keep
audience safe and
mourn their
personal losses.
News Reports: Propaganda:
Online Safety:
Obituaries/Missing Persons:
95. How does the audience engage
in adverse scenarios?
95
! The most popular public
figures were not necessarily
ones most mentioned by the
Content Producers.
! Different Spikes in
Audience's posting and
Content Producers’.
96. Attributes Most Active Audience
96
! We discover common traits of the most active
audience members.
Identify highly active
audience members.
Characterize highly
active audience
Use Mean Shift to cluster active
audience
97. What are the traits of the most active
audience members?
97
Drug Cartel Savvy
• Majority of comments about the drug cartels.
• Produced the most and longest comments.
• 33% of the most active.
• Most references to locations.
Geographers
• No reference to any public figures but did reference geographical
locations.
• Produced the second-most and longest comments.
• 1% of the most active.
Government Gossipers
• Produced the least comments and shortest.
• 66% of the most active.
• Majority of comments were about the Government.
• Some practiced redundancy in their comments.
98. Understanding Online Audiences
Research Findings
98
! Friendly Environments
-Authors & Audiences collaborate to distribute surprising content and overcome algorithms.
-Authors tag as a strategy to harvest supportive audiences.
• Controversial Environments
-Authors use the strategy of adapting their self-presentation to their audience to obtain
more participation.
• Adverse Environments
-Authors collaborate with their audience to produce news reports and even offline
collective efforts.
-Authors have strategies to engage their audience: show solidarity by explaining how to keep
safe, use offline events to recruit newcomers.
-Audience empowered to drive the narrative of events.
•
99. Research Takeaways Understanding Audiences
99
! Content Authors and Online Audiences collaborate to
produce collective efforts.
• Content Authors use different strategies to harvest
supportive audiences.
! Relationship between Content Producers and Online
Audiences resembles Gift Economy.
100. R2: How does the audience engage in
adverse scenarios?
100
! Offline events most effective to
recruit new participants to page .
! Most popular content created
during major offline events.
101. Integrating the Audience in a Post:
People Tagging in
Posts
Tagees receive
Content shared
with
Shared with
tagees’ friends &
People Tagged in Post
Tag
101
102. Engaging Online Audiences
102
! I proposes two system designs to engage online
audiences:
(1) Authors visualize the
characteristics of their
audience
(2) Authors use knowledge to
decide themselves strategy to
engage audience & start
collaborations
(1) Authors send out automated
agents that try strategies to engage
audience & start collaborations
! Visualizing Audiences ! Automated Agents
103. Research Takeaways
Understanding Audiences
103
• Content Authors use different strategies to harvest
supportive audiences.
! Content Authors and Online Audiences collaborate to
produce collective efforts.
! Relationship between Content Producers and Online
Audiences resembles Gift Economy.
104. Research Takeaways
Engaging Audiences
104
! Multifaceted DataVisualizations help content producers to
better identify strategies (gifts) to motivate collaborations
with their audience.
! Autonomous Agents help content producers to probe
strategies to motivate collaborations with their audience.
105. Engaging Online Audiences
105
! I propose two system designs to engage online audiences in
gift economy:
(1) Content producers visualize the
characteristics of their audience
(2) Content producers use knowledge
to decide themselves strategy to
engage audience & start collaborations
(1) Authors send out automated
agents that try strategies to engage
audience & start collaborations
! Visualizing Audiences ! Automated Agents
107. Total number of participants 270
Participants recruited from FB 32
Participants recruited outdoors 88
Participants recruited from
Amazon Mechanical Turk
150
Sex Demographics 43% Female, 57% Male
Age Demographics 18-68 years old, median age of 22
Participant Demographics
107
108. Can Online Bots be used to
engage online audiences for
a content producer’s cause?
111. Takeaways
• Need for spaces where people can collaborate to design their
online image and distribute meaningful content.
• Need for spaces to organize large audiences for real world
activities
• People assume roles.
• Need for spaces where people can lend their identity for a
collaboration.
• People want to reach and meet new strange audiences. Need to
provide them with the bridges.
• People have to invest time in manually learning about their
audiences.
111
113. Engaging Online Audiences
113
! I propose two system designs to engage online audiences in
gift economy:
(1) Authors visualize the
characteristics of their
audience
(2) Authors use knowledge to
decide themselves strategy to
engage audience & start
collaborations
(1) Authors send out automated
agents that try strategies to engage
audience & start collaborations
! Visualizing Audiences ! Automated Agents
114. R3: What are the traits of the
most active audience members
in adverse scenarios?
115. Results
Frequencies Participants reported to being:
a) taggers; b) tagees; c) viewers.
115Taggers Taggees Viewers
0
10
20
30
40
50
Percentage
1 2 3
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Nothing
Little
Somewhat
Much
116. Results
How much Participants reported to enjoying being:
a) taggers; b) tagees; c) viewers.
116
Taggers Taggees Viewers
0
10
20
30
40
50Percentage
1 2 3
−1
−0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
Not at all
Very little
Somewhat
To a great extent
117. Takeaways
• Audiences in general enjoy greatly being
involved in content.
• Content creators in general enjoy the interaction
somewhat, likely due to stress of integrating
people in uninteresting/ inappropriate content.
117
118. Understanding Online Audiences
Research Takeaways
118
! Content Authors and Online Audiences collaborate to
produce collective efforts.
• Content Authors use different strategies to recruit and
engage supportive audiences.
119. What type of self-presentations get
more responses from an audience?
{Person’s Post Self
Presentation
(Implicit Interests)
{Person’s Profile Self
Presentation
(Explicit Interests)
119
120. R2: How does the Audience
engage in Adverse Scenarios?
122. R3: What are the traits of the most active audience members?
Geographers
• No reference to any public figures but did reference geographical
locations.
• Produced the second-most and longest comments.
• 1% of the most active.
123. R3: What are the traits of the most active audience members?
Government Gossipers
• Produced the least comments and shortest.
• 66% of the most active.
• Majority of comments were about the Government.
• Some practiced redundancy in their comments.
124. LiveJournal Data
•
Collected LiveJournal (LJ) Posts from ONTDP and Profiles from
all bloggers to ONTDP from March 30th 2012 to July 11th 2012.
Authors 296
Commenters 1,972
Interviewees 12
LJ posts 1,200
Comments 30,934
Profile Tags 9,812
Post Tags 1,622124
126. Results
• People who tried to match profiles with posts
were less in touch with audience: got less
comments. 126
People who got most
comments from their
audience.
127. Takeaways
• Support Author’s and Audience’s Role Play.
• Help Authors and Audiences visualize the interests
and role play of others.
• Need for Socially Aware Presentation Cards.
127
128. Talk Outline
• Integrating the Audience (People Tagging)
• Self Presentation and Audience Participation
• Tools for Targeting Audiences
• Audiences in Adverse Scenarios
Friendly
Adverse
128
129. R1: What do Content Producers
Share with their Audience in
Adverse Scenarios?
130. Participatory Media Content Distribution
130
Community
Audience
Reporter Publisher
Reporter
Advertiser
Editor
Community
Spreadable Media, H. Jenkins, et al.
"We the Media”, Dan Gilmor
Emerging Relationships, Tensions,
Experiences … are Unclear!
131. 131
! Online audiences engage with content similar to
players in Alternate Reality games and fandoms.
! Relationship between content producers and
online audiences resembles Gift Economy.
Understanding Online Audiences
132. Participatory Media Ecosystem
132
Blogsphere: the emerging Media Ecosystem by
John Hilter, Microcontent News
Content Producers
Radio,TV, Newspaper, UsersSources
Story Ideas
Story Iterations
Conversation
Audience
Audience
133. Many tools to analyze audiences …
but useful to engage?
134. Hard to Design Engaging Tools
134
! Most platforms are adaptations of traditional marketing
tools, ignoring emerging dynamics.
*R. Zamora, Individual Report on
‘‘Audience Interactivity and Participation’’,
*A. Bergström,Audience Interactivity and
Participation, 2012.
Tools measure the wrong things
Use one size fits all reporting.
Most ignore text analytics.
User needs to act as detective!
135. Understanding Online Audiences
Research Takeaways
135
! Creators & Audiences struggle with algorithmic filtering.
! Creators & Audiences collaborate to overcome algorithms.
! Creators study their audiences to harvest support for
different collaborations.
! Creators probe different strategies to harvest supportive
audiences for collaborations.
! Audience empowered to define collectively a narrative
without following author lead.
137. Understanding Online Audiences
Research Takeaways
137
! Authors and Online Audiences interact with each other
to produce collective efforts.
! Authors use different strategies to recruit and engage
with supportive audiences.
138. Understanding Online Audiences
138
Research Contributions
! Creators & Audiences decide to collaborate to popularize content,
overcoming sometimes even algorithmic filtering.
! Creators study their audiences to identify strategies to harvest support for
different collaborations.
! Creators probe different strategies to harvest supportive audiences for
collaborations.
! Audience empowered to define collectively a narrative without following
author lead.
139. 139
Previous research considers that information
becomes “viral”. Removing decision from
people.
Understanding Online Audiences
140. Understanding Online Audiences
140
Research Findings
! Creators & Audiences decide to collaborate to popularize content,
overcoming sometimes even algorithmic filtering.
! Creators study their audiences to identify strategies to harvest support for
different collaborations.
! Creators probe different strategies to harvest supportive audiences for
collaborations.
! Audience empowered to define collectively a narrative without following
author lead.
143. Understanding Online Audiences
143Friendly Adverse
! I use social media to understand the experiences,
relationships, tensions, and interactions emerging from
content producers and their online audience.
Burbank's (1967) Interactive audience space,
Berkenkotter C (1981) Understanding a writer’s awareness of audience
144. Impact of this work
! This research expands our understanding of audiences
and content producers in wider spectrum.
! Shifts design of tools to engage online audiences from
market economy to gift economy.
144
Jenkins, H. "Interactive audiences? The collective intelligence of media fans.
Baym, N. et al., "Amateur experts International fan labour in Swedish independent music."
146. Engaging Online Audiences
146
! I use the understanding to design novel tools that help
authors to better engage with their online audience.
147. Engaging Online Audiences
Design Proposals
147
! Human in the loop interfaces to
target audiences.
! Multifaceted data visualizations to
help creators target audiences for
their different collaborative tasks.
! Systems that let creators probe
different strategies to recruit and call
audiences to action.
148. 148
Research Contributions
! Analyses of Online Audiences
! Qualitative and quantitative analysis of audience targeting
mechanisms (Hypertext 2015)
! Analysis of content producer self-presentation and audience
engagement (work led to patent submissions with Intel)
! Design and Evaluation of Tools for Engaging Online
Audiences.
! [Hax]
! [CSCW submission]
150. (1) Authors
visualize the
characteristics of
their audience
(2) Authors use
knowledge to decide
themselves strategy to
engage audience & start
collaborations
(1) Authors send out automated
agents that try strategies to engage
audience & start collaborations
! Visualizing Audiences ! Automated Agents
154. Challenges
154
! Difficult to design engaging media.
*Pasquali, N,“Emerging Topics in the Research on Digital Audiences and
Participation”
*H. Sanchez Gonzales, Connectivity between the Audience and the Journalist
Source: Mashable
155. Understanding Online Audiences (Adverse)
Savage S., Monroy-Hernandez A., Participatory Militia, CSCW’15
Participatory Militias
An Analysis of an Armed Movement’s Online
157. Goal
Use the “Courage For” pages as a lens to understand
1) the type of online content shared by content
creators in adverse scenarios;
2) how the online audience engages with content
creators;
3) characteristics of the most active audience
members.
166. 166
Evaluation
Launched our tool to the public. Study type of audience recruitment for which our tool is
used, and successes. Interviews and surveys to understand subjective perceptions.
167. 167
Results
Multi-modal visualizations facilitated:
-Serendipitous Discoveries
-Visualizing Other People’s Likelihood of Participation (Geo & Knowledge-based)
-Visualizing Diffusion (Spread Information)
-Audience Diversity (Cultural and Tastes)
-Audience Verification
169. • Context Aware Systems for
Opportunistic Participations
Supporting Online Audiences
Savage S., et al., I'm Feeling LoCo: A Location Based Context Aware Recommendation
System, Lecture Notes in Geoinformation and Cartography. Springer.
170. Goals
A tool that facilitates volunteering & contributing
opportunistically:
• Understands users’ lifestyle and preferences.
• Understands users’ current context (activity)
• Match tasks to available and interested users.
170
171. 171
System Design
Savage S., et al., I'm Feeling LoCo: A Location Based Context Aware
Recommendation System, Lecture Notes in Geoinformation and Cartography. Springer.
172. 172
Evaluation
Savage S., et al., I'm Feeling LoCo: A Location Based Context Aware
Recommendation System, Lecture Notes in Geoinformation and Cartography. Springer.
Launched our tool to the public. Study opportunistic participations
that our tool facilitates. Interviews and surveys to understand
subjective perceptions.
173. We identified posts that referenced the public figures
and organizations involved in the conflict.
Content Analysis: Public Figures
1. Collected Wikipedia & Proceso
articles related to conflict in
the region
2. Identified all proper names. 3. Add or merge alternate
names for
each public figure.
“Estanislao Beltran”,
“Papa Smurf”,
“ Estanislao”
Public Figure 1
Public Figures 1,2
Public Figure 5
Public Figure 3
Public Figure 1
Public Figure 1
4. Identified VXM posts
and comments that
mentioned each public
figure.
175. Grounded Theory
StagePurpose
Codes: Identifying anchors that
allow the key points of the data to be
gathered
Concepts: Collections of codes of
similar content that allows the data to be
grouped
Categories: Broad groups of
similar concepts that are used to generate
a theory
Theory: A collection of explanations
that explain the subject of the research
176. Related Work
Self-Presentations to Audiences
• I tweet honestly, I tweet passionately: Twitter users, context collapse, and the
imagined audience, 2011, Marwick A. , boyd D.
• Managing impressions online: Self‐presentation processes in the online dating
environment, 2006, N Ellison, R Heino, J GibbsPal, A., and Counts, S. 2011. What’s in a @name? How name value biases judgement of
microblog authors
A familiar face(book): profile elements as signals in an online social network,
2007, Lampe, C.; Ellison, N.; and Steinfield, C.
Lampe, C.; Ellison, N.; and Steinfield, C. A familiar face(book): profile elements as
signals in an online social network.
Mining Analytics
Mining expertise and
interests from social
Collective Intelligent
Augmenting Human Intellect, 1962, Engelbart Douglas,
Collective Intelligence: Mankind's Emerging World in Cyberspace.
1999, Pierre Levy.
177. Related Work in Author’ Self Presentation
and Audience
177
• Marwick A., boyd d., I tweet honestly, I tweet
passionately: Twitter users,
context collapse, and the imagined audience, 2011, New Media and Soci
• Ellison N, Heino R, Gibbs J, Managing
Impressions online: self presentation processes in the online dating enviro
2006, Journal of Computer‐Mediated Communication
• Lampe, C.; Ellison, N, and Steinfield, C, A familiar
face(book): profile elements as signals in an online social network, 2007,