The document discusses the challenges of visualizing complex systems and presents an ongoing project called SystemViz that aims to address these challenges. It notes that complex systems are difficult for humans to understand without visual aids due to limits in working memory. The project involves taking stock of existing visualization methods, identifying concepts that are not easily depicted, and exploring recurring visualization problems. The overall goal is to develop guidelines and tools to help design effective visualizations for understanding complex systems.
1. D E S I G N + V I S U A L T H I N K I N G
V . 0 . 5 - B B Y P E T E R S T O Y K O
THE CHALLENGE
Complex systems are difficult to understand without the
aid of visuals. There are too many moving parts to mentally
keep track of. The parts interact in too many ways. The whole
system is cognitively overwhelming insofar as it cannot be
absorbed in one go without the aid of an external reference.
That is partly due to humans' inability to juggle more than
a few complicated ideas in working memory at one time.
Thus, visuals are a simplifying and organizing device that
complements the way human naturally think if they are
designed well. This poster is an early glimpse of a larger
project (called SystemViz) that explores what it means to
design such visuals well.
TAKING STOCK OF VISUAL METHODS.
The first challenge is that there are many
ways to display systems visually. Diagram
formats and notations have proliferated.
Some correspond to particular disciplines
or tasks. Each has innumerable minor
variations. It is worth taking stock of these
"visual vocabularies" because the choice of visualization
method skews the way systems are understood. Different
methods emphasize different things and removes certain
details from the picture: all are, by definition, a form of
reductionism; all contain implicit assumptions about what a
system is and how it behaves. For example, one method may
depict systems as stocks and flows, another may depict
them as logical sequences; another as causes, effects,
and two-way dynamics; another as signals and boundaries.
These differences depend on the underlying theory in use,
but also the unthinking tendency to adopt a visualization
method out of convenience or habit. The result is a paradox:
without visuals to help model systems, understanding
is limited; visual conventions frame the way systems are
understood, which can undermine understanding. Thus,
the first task of this project is to provide an inventory
of the various visualization methods by grouping them
according to their main points of differentiation. This is a
first approximation. The aim is to solicit feedback about this
way of organizing visuals and determining if there are major
errors of omission. The aim is also to have a discussion of
the biases and blind spots of the respective methods.
IDENTIFYING MISSING CONCEPTS.
There are many systems concepts that
are applicable across practical and
academic disciplines. The overall literature
is wide and deep. Some concepts are
wide recognized; others less so, partly
because of a lack of trans-disciplinary
engagement. That presents an opportunity for visuals to act
as an interdisciplinary lingua franca. That also raises some
important questions: What systems concepts are not easily
depicted using conventional methods of visualization?
Identifying these concepts is the second aim of the project.
The hope is to start a conversation about visualization
methods that might fill these glaring omissions.
EXPLORING VISUALIZATION CHALLENGES.
Creating an inventory is an opportunity to
reflect on recurring problems and trade-
offs. That is the third task of the project.
The goal is to provoke a discussion about
ways of overcoming these difficulties.
A few lessons from the discipline of
information design provide partial guidance. The ultimate
goal is to turn these stock-taking exercises and "guide
posts" and turn them into a design and teaching tool (a
"codex") in an appealing and accessible form. This post is an
early step in that direction. Inspiration is drawn from David
Garcia's Manual of Architectural Possibilities series of road-
map-like foldable posters. Interactive, screen-based versions
may also emerge. If the project yields some interesting
techniques, the goal is to make those openly available to
members of the larger community interested in systems
thinking.
Peter Stoyko is an interdisciplinary social science and
information designer at Elanica, a management consultancy
that specializes in service design and governance.
Feel free to contact him at peter.stoyko@elanica.com or visit
the project page at systemviz.elanica.com.
?
NO-NAME CONNECTIONS
Unspecified links between items make
relations and dynamics ambiguous,
often suggesting analytical evasion.
Link labels, badges, symbolic ends, or
encodings (e.g. color coding) indicate
nature of link.
// FIX //
VISUALIZATIONHURDLES
VISUALIZATIONTYPES
SPAGHETTI TANGLE
Unmanageable links between objects
makes it difficult to distinguish
relationships.
Prioritize and differentiate lines.
Use layering techniques. Add visual
affordances to guide the eye.
// PARTIAL FIX //
STATE EXPLOSION
Unmanageable numbers of diagram
nodes, especially for number of
potential states in Petri nets.
Increase level of abstraction at the
cost of precision. Use different levels
of analysis for crucial details.
// TRADE-OFF //
EVERYTHING CONNECTED
The temptation to connect everything
to everything else because, at some
level, a relation can be imagined.
Rank and differentiate relationships
according to well articulated analytical
priorities.
// FIX //
FALSE EQUIVALENCE
Failure to sufficiently differentiate
linkages creates mistaken impression
of equal importance.
Differentiate lines in ways that convey
magnitude and qualitative differences.
Add text labels and numerical weights.
// FIX //
NOTATION OVERLOAD
Number of symbols and distinctions
exceeds viewer's ability to differen-
tiate and juggle in working memory.
Reduce number of encodings,
especially for minor distinctions. Do not
rely on legends.
// PARTIAL FIX //
PERCEPTUAL SIMILARITY
Basic shapes used in diagram are
not sufficiently differentiated so that
important distinctions are overlooked.
Use shapes (and other encodings,
such as color and text style) with
higher levels of contrast.
// FIX //
ETHNOCENTRIC SYMBOLISM
Symbolic visual objects (such as
icons) carry cultural meanings that are
not universally understood.
Research cultural background of
symbols. Use symbols recognizable
across cultures.
// PARTIAL FIX //
SUBJECT MISMATCH
Diagram method is poor fit to the
nature of the subject matter because
of mismatched theory of use.
Use more suitable method. Abandon
highly standardized notation for more
bespoke graphics.
// FIX //
NEGLECTEDCONCEPTS?
A
C
E
E
L
J
J
Z
A
M
C
4
5
6
7
M
4
5
6
Perforated Collection Tube
Fresh Water
Reverse Osmosis Membrane
Pressure Vessel
Concentrate
Fresh
Water
Feed
Water
Fresh
Water
LINKS BETWEEN OBJECTS SHOWING CAUSAL
DIRECTION, MOVEMENT, OR SEQUENCE
FLOW TRANSITIONPETRI NETS AND OTHER MATHEMATICAL GRAPHIC
NOTATIONS SHOW SYSTEM PHASES WITH PRECISION
LOOPSTOCKS, FLOWS, AND SYSTEM DYNAMICS ARE
MAPPED AS LOOPS
SITUATESYSTEM DIAGRAMS ARE LAYERED OVER MAPS
OF PHYSICAL SPACE AND OTHER DIAGRAMS
ARRANGENODES AND LINKS ARE ARRANGED AND CATEGORIZED
WITH VARIOUS SPATIAL ARRANGEMENTS
EXPLAINILLUSTRATED DIAGRAMS SHOWS HOW A SYSTEM
WORKS BY MIMICKING REAL-WORLD APPEARANCE
Geothermal Heat
Magma
Condenser
Steam
Hot Air &
Water Vapor
Waste Water
Groundwater
Replenishment
Top-up
Injection
Well
Production
Well
Impermeable Rock
Cooling Water
Impermeable Rock
Conductive RockConductive Rock
Groundwater
Infiltration
High Lands
Water
Run-off
Warmed Water
Cooling
Water
Power
Transmission
Power Generation
Power
Transmission
Rain Water
Coastal Winds
Pump
Turbine
Generator
Water
Treatment
Cooling
Tower
Households
Transformer
Condensate
Heated Aquifer
LIKENVISUAL ANALOGIES COMPARE SYSTEMS TO
OTHER SYSTEMS THROUGH BLENDING
EXPOSEPART AND WHOLE RELATIONSHIPS OF SYSTEM ARE SHOWN
REALISTICALLY THROUGH CUT-AWAY AND EXPLODED VIEWS
YES
NO
NO
Motion
detected Animal?
Human?
Guard?
Isolate
Individual
Remove
Individual
Escort to
Office
Send to
Office
On Duty?
Machine?
Beacon? Alone?Record
Number
Valid?
Danger?
Isolate
Machine
Escort
to Lab
Guard
Check
Continue
monitoring
YES
NO
NO
NO
NO
YES
YES
YES
YES
NO NO
YES
NO
YES
YES
Report
Incident
Input
Reason
Sound
alarm
Remove
Machine
G E N E R I C N O T A T I O N
PROCESS
DECISION
DELAY
PROCESS FLOW
TERMINAL
DATA
PROCESS GROUP
DISPLAY
MANUAL INPUT
TRANSFER
MANUAL LOOP
LOOP LIMIT
CONDITION
ORGANIZATION
PREPARATION
STORED DATA
DOCUMENT
PARALLEL MODE
SUB-PROCESS
CONNECT
SORT
MERGE
COMMENT
AND
Text
CHECK
COLLATE
TO PAGE
OR
JUMPREFERNCE
Intelligence
Gathering
Product
Specifications
Prototype
Refine
User
Testing
Tester
Screening
Test-site
Preparation
Outreach
Implement
Participant
Recruitment
Data
Requirements
Field
Research
Findings
Reviewed
Advisory
Review
Study
Prepared
Finalization
Client Needs
Assessment
Client
Feedback
Stakeholder
Mapping
Product
Launch
Participatory
Research
R1
Integrative
Learning
B1
Product
Development
R2
Knowledge
Mobilization
R3
Iteration
R1b
Iteration
R2b
Testing
R3c
Tweak
R2c
( Arita, 49 )
( Ito, 21 )
( Nishi, 31 )
( Fukuyama, 42 )
( Takeuchi, 38 )
( Enomoto, 26 )
( Sakoda, 55 )
Request End
Submission
Process
Flag
Expedited
Submission
References
Checked
Record Ready
Request
Routed
Request
Filed
Record
Reviewed
Add More
References
Process
( Matsuki, 22 )
CAUSAL
LINK
POSITIVE
EFFECT
DYNAMIC
LINK
NEGATIVE
EFFECT
LOOP
LABEL
BOUNDARY
DELAYED
EFFECT
FLOW
OUTFLOW
BIFLOW
A2
BALANCED
LOOP
(OR )B
REINFORCING
LOOP
(OR )R
FLOW
RATE
OR
OR
G E N E R I C
N O T A T I O N
STATE
STATUS OF
SYSTEM ITEM
PSEUDO STATE
EXOGENOUS
STATE
SEE SCHOLARPEDIA ARTICLE
INSERT
TRANSITION
CHANGE TO
NEW STATE
SUBSYSTEM
MULTIPLE PARTS
TO STATE OR
TRANSITION
DELAY
IN NUMBER OF
TIME UNITS
TIME STAMP
START OF MOVE
IN TIME UNITS
TOKENS
OBJECT MOVING
THROUGH SYSTEM
TOKEN LABEL
DATA SEPARATED
BY COMMA
@+3
3
G E N E R I C
N O T A T I O N
M U LT I P L E L A Y E R S
L A Y E R E D L E V E L S
O F A N A LY S I S
E X P L O D E D
C U T A W A Y
THE FOLLOWING ARE SOME
EXAMPLES OF SYSTEM
CONCEPTS THAT ARE BEING
COLLECTED AS PART OF
THE PROJECT TO SEE IF
EXISTING VISUALIZATION
METHODS CAN ADEQUATE-
LY REPRESENT THEM.
CURRENTLY, THE DATABASE
HAS APPROXIMATELY A
HUNDRED ITEMS GATHERED
FROM ACROSS NATURAL
AND SOCIAL SCIENTIFIC
DISCIPLINES.
ATTRACTOR
An element that draws
others towards it or
distorts its trajectory
through an attractive force.
REPELLER
An element that pushes
others away or distorts its
trajectory through repelling
force.
SIDE EFFECTS
Externalities or unanticipated
consequences caused by
a force or dynamic between
two objects.
DISTAL DRIVER
A background cause observ
able at a high level of analysis
("bigger picture"); a diffuse
factor acting on more obvious
factors in a causal chain.
KLUDGE
A make-shift work-around
to cope with an immediate
problem, adding variation
(often complexity) to system;
source of "jagaad" innovation.
ENABLER
A factor that gives another
factor a boost, either as a
necessary ingredient or as a
contributing one.
PROXIMATE DRIVER
A variable that directly
causes another variable to
change; that is, an immediate
cause.
FRICTION
A factor that impedes
another factor directly to
blunt, delay, or skew.
An obstacle that prevents
a variable from running its
course by affecting other
variables in a causal chain
PROTECTOR
A blocker that preserves the
state of a variable from some
or all outside influences
Boundaries Signals
CONTAINER
A boundary that groups
items discretely by
separating them from
others
SEPARATOR
A boundary that separates
or differentiates two
groups or areas
SEMI-PERMEABLE
A boundary that blocks
certain factors but not
others depending on their
qualities
BOTTLENECK
A boundary that limits the
throughput of forces
through an area or process
DIVERSITY BETWEEN
The variation between
different categories of object
DIVERSITY WITHIN
The amount and quality of
variation found within a single
category of object
Factor that has to be present
for one thing to cause another
to change, perhaps part of the
environment
PERSPECTIVE
How the system changes
depending on the vantage
point or qualities of the
viewer.
SENSING
Parts of the system that
receive signals from the
external environment or
elsewhere in the system.
INPUT/OUTPUT
Parts of the system that take
in or expel factors from the
outside.
COMPATIBILITIES
Aspects of a system that are
able to interact functionally
or to mutual benefit because
of compatible design features
PARASITE
An outside agent that derives
benefit from an host or
system while imposing costs
Relations DomainsForces
PROCESSCONCEPT DATABASE
Sort
Store
Store
Form
Status
No MergeNew?
B
A
A
X
B
Sort
Start
Sort
Pump
X
Archive
Activity
Match
Report
Inspect
Soya
G
HJ
L
K
AM
N
F
Water
Land
Market
Soil
Sun
C
G1
C2
P2
P1
C3
G2J1
K1
2
J2
P
A
B
M2