A talk given to the "Social.Path" workshop at the University of Surrey, June 2014.
It is well established that many human abilities are context-dependent, including: language, preference judgement, memory, reasoning, learning and perception. This is usually taken as a negative – that there will be limits on our understanding and modelling of these abilities. However, what is not always appreciated is that context-dependency can be a powerful tool in social coordination and communication. This paper pulls together several theories about the cognition of context, and presents a computational model of context-dependency. It then sketches its role in social communication, coordination and embedding. It looks at some of the approaches to dealing with context in the computer science and social science literature and concludes that none of these squarely faces the problem of context dependency. This points towards a substantial gap in the research and hence a future programme.
1. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 1
The Sociality of Context
Bruce Edmonds
Centre for Policy Modelling
Manchester Metropolitan University
2. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 2
Talk Outline
1. A bit about Context
2. Context-Dependency in Cognition
3. Context-Dependency and Society
4. An Architecture for Cognitive Context
5. Learning/Identifying Context
3. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 3
Context
• “Context” is used in many different senses
across different fields
• The senses and concepts herein come from
discussions and papers presented at the
international series of conferences on
“Modelling and Using Context”
• Somewhat of a “dustbin” concept resorted to
when more immediate explanations fail (like
“complexity”!)
• Problematic to talk about, as it is not obvious
that “contexts” are identifiably distinct
4. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 4
A (simplistic) illustration of context from the
point of view of an actor
5. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 5
Situational Context
• The situation in which an event takes place
• This is indefinitely extensive, it could include
anything relevant or coincident
• The time and place specify it, but relevant
details might not be retrievable from this
• It is almost universal to abstract to what is
relevant about these to a recognised type
when communicating about this
• Thus the question “What was the context?”
often effectively means “What about the
situation do I need to know to understand?
6. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 6
Cognitive Context (CC)
• Many aspects of human cognition are context-
dependent, including: memory, visual perception,
choice making, reasoning, emotion, and language
• The brain deals with situational context effectively,
abstracting kinds of situations so relevant
information can be preferentially accessed
• The relevant correlate of the situational context
will be called the cognitive context
• It is not known how the brain does this, and
probably does this in a rich and complex way that
might prevent easy labeling/reification of contexts
7. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 7
The Context Heuristic
• The kind of situation is recognised in a rich,
fuzzy, complex and unconscious manner
• Knowledge, habits, norms etc. are learnt for
that kind of situation and are retrieved for it
• Reasoning, learning, interaction happens with
respect to the recognised kind of situation
• Context allows for the world to be dealt with by
type of situation, and hence makes
reasoning/learning etc. feasible
• It is a fallible heuristic…
• …so why do we have this kind of cognition?
8. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 8
An Illustration of the Heuristic
Model
Factor 1
Factor 2
Factor n
Factor n+1
Factor n+2
Etc.
Inferences/
predictionsModel
Factor 1
Factor 2
Factor n
Factor n+1
Factor n+2
Etc.
Outcomes
Foreground
features
Later recognition
1. Learning Situation 2. Application Situation
Possible abstraction to a ‘context’
9. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 9
Social Intelligence Hypothesis
• Kummer, H., Daston, L., Gigerenzer, G. and Silk, J. (1997)
• The crucial evolutionary advantages that
human intelligence gives are due to the
social abilities it allows
• Explains specific abilities such as imitation,
language, social norm instinct, lying,
alliances, gossip, politics etc.
• Social intelligence is not a result of general
intelligence, but at the core of human
intelligence, “general” intelligence is a side-
effect of social intelligence
10. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 10
An Evolutionary Perspective
Social intelligence implies that:
• Groups of humans can develop their own
(sub)cultures of technologies, etc. (Boyd and
Richerson 1985)
• These allow the group with their culture to
inhabit a variety of ecological niches (e.g.
the Kalahari, Polynesia) (Reader 1980)
• Thus humans, as a species, are able to
survive catastrophes that effect different
niches in different ways (specialisation)
11. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 11
Implications of SIH
• That different complex “cultures” of knowledge,
skill, habit etc. are significant
• An important part of those cultures is how to
socially organise, behave, coordinate etc.
• One should expect different sets of social
knowledge for different groups of people
• That these might not only be different in terms
of content but imply different ways of
coordinating, negotiating, cooperating etc.
• That these will relate as a complete “package”
to a significant extent
12. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 12
Social Embedding
• Granovetter (1985)
• Contrasts with the under- and over-socialised
models of behaviour
• That the particular patterns of social
interactions between individuals matter
• In other words, only looking at individual
behaviour or aggregate behaviour misses
crucial aspects
• That the causes of behaviour might be spread
throughout a society – “causal spread”
• Shown clearly in some simulation models
13. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 13
Co-Development of Shared Context
• Over time certain kinds of situation get recognised as
similar by individuals
• These individuals may further distinguish it as a specific
context
• Knowledge is remembered and stored associated with
that context
• This recognition facilitates the development of a set of
shared habits, norms, knowledge, language etc. that is
specific to that context
• The more this happens the more distinctive that kind of
situation becomes and hence more recognisable by
newcomers
• This self-reinforcing loop can entrench the context and
ensure its persistence and identifiability over time
14. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 14
Social Context
• Heavily entrenched contexts may become
institutionalised in terms of infrastructure,
training etc. (e.g. how to behave in lectures)
• Thus easily distinguishable and widely
recognised
• Socially entrenched contexts can be
explictly labelled and talked about
• This reification allows the contexts to be
further entrenched as it might be demarked
and institutionally protected
15. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 15
Social Advantages of an Ability to
(Co)Recognise Context
If (broadly and roughly) the same contexts are
recognised by different people then…
• Communication is easier – assumptions
about situation likely to be shared
• Terms specific to that kind of situation can
be developed focusing on the foreground
• Different ways of coordinating can evolve
for each different kind of situation, allowing
the kind of coordination to be suitable for
the situation
16. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 16
Some Ways in Which Dealing with
Context is Avoided by Researchers
• Some seek to only consider behaviour that
is general (i.e. across context), treating the
unexplained variance as noise
• Some retreat to only considering behaviour
within very specific contexts
• Some treat context-specific behaviour as if
it is “just more detail”, and so keep to
abstract representations to avoid this
• Others simply ignore it and hope that more
generic representations will do
17. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 17
Basic Cognitive Model
• Rich, automatic, messy, unconscious
context recognition using many inputs
(including internal ones) – ML
• Crisp, conscious, explicit processes using
material indicated by cognitive context:
reasoning and learning – AI
Context
Recognition
Context-Structured
Memory
Reasoning/Plan
ning
Belief
Revision/Learnin
g
18. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 18
Context-Dependency Makes Learning
and Reasoning Feasible
• Context allows us to progress beyond the ‘loose’ loop of:
repeat
learn/update beliefs
deduce intentions, plans and actions
until finished
• to a more integrated loop of:
repeat
repeat
recognise/learn/choose context
induce/adapt/update beliefs in that context
deduce predictions etc. within that context
until predictions are consistent…
and actions/plans can be determined
plan & act
until finished
19. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 19
context
identification
CIS
CDM
inputs from
enviroment
or problem
actions
and/or
plans
negative feedback if
wrong context
feedback on under- or
over-determination
within context
update or
learning
cognitive
context
LL
IS
within
context
reasoning
An Implemented Cognitive Architecture
Local Learning
algorithm
Context-
Identification
System
Context-Dependent
Memory
Inference
system
20. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 20
Example ABM using this Architecture
• Artificial Stock Market with market maker and trading costs
• Several traded stocks and cash
• Fundamentals in form of slow random walking dividend
rate for each stock
• Agents learn models to predict strategies using a rich GP-
like language
• Models evaluated over recent time period, then best
selected, effect of actions predicted and action selected
• Memory a space of market volume vs. volatility where
models are ‘placed’ as learnt
• Context agents additionally select a region of memory
according to current conditions and learn/evaluate/select
models according to the currently region of memory
• Other agents the same with the whole memory selected
21. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 21
0
5000
10000
15000
20000
25000
30000
0 100 200 300 400 500
Time
TotalValueofAssetsTotal Assets in a Typical Run
Black=context, White= non-context
22. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 22
Example – models in the cognition of
a trading agent
700
750
800
850
900
950
750 850 950
Volume - past 5 periods
Volatility-past5periods
23. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 23
The model contents in snapshot of
one trader
model-256 priceLastWeek [stock-4]
model-274 priceLastWeek [stock-5]
model-271 doneByLast [normTrader-5] [stock-4]
model-273 IDidLastTime [stock-2]
model-276 IDidLastTime [stock-5]
model-399
minus
[divide
[priceLastWeek [stock-2]]
[priceLastWeek [stock-5]]]
[times
[priceLastWeek [stock-4]]
[priceNow [stock-5]]]
24. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 24
Other Useful Loops
• If there is not enough knowledge to determine
action, expand the context scope
• If action is over-determined then narrow the
scope
context
identification
CIS
CDM
inputs from
enviroment
or problem
actions
and/or
plans
negative feedback if
wrong context
feedback on under- or
over-determination
within context
update or
learning
cognitive
context
LL
IS
within
context
reasoning
25. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 25
Learning/Recognising Context
• Context seems to be recognised in a rich,
unconscious and “fuzzy” manner
• Using both external (how others behave)
and internal (emotions, goals) cues
• Some social contexts will have labels but
others only indirectly inferable
• However, social context should be
identifiable by clusters of behaviours,
norms, habits etc. all having the same
conditions of occurence
26. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 26
• A possible context is where models that fit different kinds of behaviour have a
similar scope
• You know you are in the wrong context if many aspects of your knowledge fail
simultaneously
Clusters of Conditions of Application
suggest a Context
M1 M2
M1
suggests a context
Spaceofpossible
conditions
27. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 27
Cleveland Heart Disease Data Set – the
processed sub-set used
In processed sub-set:
• 281 entries
• 14 numeric or numerically coded attributes
• Attribute 14 is the outcome (0, 1, 2, 3, 4)
• Some attributes: age, sex, resting blood
pressure (trestpbs), cholesterol (chol),
fasting blood sugar (fbs), maximum heart
rate (thalach), number of major vessels (0-
3) colored by flourosopy (ca)
• From the Machine Learning Repository
28. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 28
Fitting a Global Model (R=56%)
Num = -0.01*age + 0.17*sex + 0.20*cp + 0.00*trestbps + 0.10*restecg + -
0.01*thalach + 0.23*exang + 0.18*oldpeak + 0.16*slope + 0.43*ca + 0.14*thal + -
0.60 (+/- 0.83)
29. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 29
Looking for Clusters in HD Data Set
(Start of Process)
30. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 30
Final Set of Clustered Solutions
• Final solution
set after some
time.
• Still complex but
some structure
is revealed
• Note presence
of “fbs” despite
not being
globally
correlated and
that “chol”
helped define
the context
space
31. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 31
So What?
• Human behaviour might be more regular when
situations are divided into contexts
• Identifying similar context may gives clues as
to when regularities can be generalised
• Context-sensitivity allows different kinds of
coordination for different kinds of situation
• Big Data may allow for clues to context
• Accepting the complication of context may
facilitate a better interface with the social
sciences
• It may allow for better human-machine
interfaces (ones that are less socially inept)
32. The Sociality of Context, Bruce Edmonds, Social.Path, Surrey, June 2014, slide 32
The End
Bruce Edmonds
http://bruce.edmonds.name
Centre for Policy Modelling
http://cfpm.org
These slides will be uploaded to:
http://slideshare.net/BruceEmonds
Hinweis der Redaktion
AI, NL, Sociology, Philosophy, Mobile devices, Psychology, Cognitive Science
For detailed argument seem my previous papers on this
Dustbin Like complexity
will talk about this problem later
Imagine a professor of physics in a wild place – does his intelligence help him to survive?