Towards a Context-Sensitive Structure for Behavioural Rules

siennaredwoodΤεχνίτη Νοημοσύνη και Ρομποτική

23 Φεβ 2014 (πριν από 3 χρόνια και 1 μήνα)

45 εμφανίσεις

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
1


Towards a Context
-
Sensitive
Structure for Behavioural Rules

(
Context, Scope, Antecedents,
and Results)

Bruce Edmonds

Centre for Policy Modelling
,

Manchester Metropolitan University

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
2


Summary of Talk:
a view from
Cognitive Science

Suggest dividing behavioural rules into 4 bits:


Context


Scope


Antecedents


Results


Since this, I argue, seems to align with human
cognitive structure


Which are each dealt with and updated in
different ways (making their use feasible)


And thus might be a more “natural” structure
for human behavioural rules

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
3


Different Aspects I

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
4


Different Aspects II

Universe of Knowledge

Knowledge indicated by current cognitive context

Knowledge that is possible to
apply given circumstances

Cause1 & Cause2…


Result1 & Result2…

Cause3 & Cause2…


Result5 & Result2…

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
5


Context

Bit 1:

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
6


A (
simplistic
) illustration of context from the
point of view of an actor

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
7


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?


Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
8


Cognitive
Context (
CC
)


Many

aspects of human cognition are context
-
dependent, including: memory, visual perception,
choice making, reasoning, emotion, and language


The brain somehow deals with situational context
effectively, abstracting
kinds of situations
so
relevant information can be easily and 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


Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
9


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 with social roots in terms
of the coordination of action, norms, habits

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
10


Some Possible Examples of Cognitive
Context?


Greeting someone you do not know


A lecture


An interview


Being Lost


Being Socially Embarrassed


Travelling on a train/bus


Leaving home to go somewhere


Accidently bumping into someone you do
not know on the pavement/in the corridor

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
11


Some Research Responses to
Context
-
Dependency

A number of responses:


Only do research
within

a single context,
resisting any
generalisation



Only use discursive, natural language
approaches where context is implicitly dealt
with (but also mostly hidden)


Try to see what (inevitably weaker
knowledge) is general across the various
contexts in what is being studied

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
12


Context
-
Dependency and
Randomness

Lots of
information
lost if
randomness
used to
“model”
contextual
variation


Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
13


However


Although Cognitive Context in General
might be hard to identify


Socially Entrenched Contexts are often
rather obvious


But one needs to drop the imperative of
looking (
only
) for abstract and general
theories for
behaviour


Being satisfied with more “mundane” and
context
-
dependent accounts


Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
14


Choice and Update of Cognitive
Context


CC is largely learnt from experienced
situations in a rich and unconscious way


Occasionally one can
realise

one has the
wrong context if a lot of the detailed
knowledge it indicates is simultaneously
wrong

but which is the right CC is a matter
of recognition from past positive learning


Once CC is learnt it is very difficult to
change, but new CC can still be learnt

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
15


Identifying Context from Narrative
Evidence


Apart from socially entrenched contexts
(lectures, parties, interviews etc.)…


…the relevant CC is hard to identify from
narrative evidence because:


To a large extent, we
recognise

the right CC for
any text unconsciously and easily


The CC are learnt in a rich, “fuzzy” manner over a
long period of time by inhabiting them which resists
reification


This is one of the prime needs:
how to “mark
up” the CC behind narrative evidence?


Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
16


Scope

Bit 2:

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
17


About Scope


By “scope” I mean the reasoning as to which
knowledge is possible given the circumstances


For example, if all the seats are taken in a
lecture, then the norms, habits and patterns as
to where one sits do not apply


Reasoning about scope can be complex and is
done consciously


However once
judgements

about scope are
made then they tend to be assumed, unless
the situation changes critically

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
18


Scope vs. Cognitive Context


Both scope and cognitive context determine
which knowledge is useful for any particular
situation that is encountered


However, they play very different roles:


CC is learnt using pattern recognition over a long
time, but then is largely a ‘given’, is almost
impossible to change when learnt, is quick and
automatic and is socially rooted


Scope is largely reasoned afresh each time, taking
effort to do so, is possible to re
-
evaluate but only if
needed, and is more individually oriented

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
19


Identifying and
modelling

scope


Compared to CC, scope is relatively well
studied using formal models of reasoning


e
.g. Updating
Markoff
/state representations of
causation, non
-
monotonic logics, causation in
Baysian

networks etc.


Scope plays a relatively explicit part in human
language, sometimes being explicitly stated, at
other times using relatively well understood
rules


e.g. conversational
implicature


It is often possible to infer participant’s
judgements

as to scope and possibility, when
not explicitly mentioned

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
20


(
local
) Narrative Steps

Bits 3&4:

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
21


Encoding Narrative Steps


*If*
CC and scope is identified then,
I
hypothesize
, the local narrative structure will
be easier to understand, because changing
CC and/or scope can do a lot of the “work” in
expressing/encoding knowledge


Within

CC & scope I suggest a simple basic
structure of sets of statements of the form:

(
on the whole
)
Z

follows/followed from
A, B…


A
very

special case of this is when we say
that:
A, B…

implies
Z

or that:
A, B…

causes
Z



(I will write A, B…

Z), where A, B are the
“Antecedents” and Z is the Results


Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
22


About Narrative Steps


These might not be crisp but of the nature
More A and B tends to result in more Z


These are often chained in forwards,
branching or backwards manner to make an
inference or a narrative


(even quite classical) formal logics and
annotation systems capture these


Most AI/expert systems encode these, but
rarely touch on scope (that is advanced AI)
and never on Context

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
23


Conclusion

Towards a Context
-
Sensitive Structure for Behavioural Rules
, Bruce Edmonds, Informal Workshop on Qual. Evidence & Rules, MMU, Sept. 2012, slide
24


CSAR as a bridging structure between
narrative text and
behavioural

rules

*IF*
this structure turns out to be a useful and
“natural” encoding of human narrative
knowledge/expression then two steps are
needed:

1.
Techniques to capture/approximate/guess
appropriate Cognitive Contexts and Scope
judgments from Narrative Text

2.
AI/Computer science architectures that
make the encoding and use of CSAR
structured knowledge