Transcript Socially distributed cognition, artefacts, organisations

‘Things to think with’ in organisational action
The presentation will examine the function, use and application of
distributed cognition, focusing on its role in social and organisational
activity, and its application in HCI design. Distributed cognition is a
technique that draws from the cognitive and social sciences, utilising
ethnographic techniques of data collection to examine how information
processing can be conducted over a larger unit of activity than the
individual, incorporating multiple individuals and external tools. The use of
the framework has great potential value in the design of information
systems, and particularly, of interaction design because of its orientation
towards the co-ordination of information representations across external
media. However, the technique has practical limitations to its use and a
number of academic questions regarding its theoretical foundations, and
these will also be addressed.
socially distributed cognition
artefacts, organisations and design
mark perry
Brunel University
<[email protected]>
<people.brunel.ac.uk/~cssrmjp>
function, application and use of DCog
Theory
Application and Use
 what
 examples
distributed cognition
is…
 DCog, specifically
 theoretical background
 novel, interesting or
useful?
(data collection
and analysis)
 emerging research
 HCI and CSCW
 theoretical and practical
implications
principles and mechanisms
1.
2.
the boundaries of
cognition do not stop
with the individual
a range of mechanisms
may be used in
cognitive processes
involving symbol
manipulation

thus, cognitive
processes may:
a)
involve internal
(mental) and external
(in the world)
structures
be socially distributed
be temporally
distributed
b)
c)
some background

several versions, appropriations and uses of the term
 Shared/ social/ group cognition, ecological psychology, situated cognition,
situated action/ ethnomethodology, etc.
 all dealing with ‘embodiment’: mind, body and world interacting and
influencing one another

today:
 focus on Ed Hutchin’s work (let’s call it DCog)

dear old Donald (on cognitive artefacts):


“…artefacts do not actually change an individual’s capabilities. Rather,
they change the nature of the task performed by that person. When
the information and processing structure of the artefact is combined
with the task, the result is to result is to expand and enhance cognitive
capabilities of the total system of human, task, and artefact”
why the interest from HCI?
1. making the jump: easy for cognitive scientists to accept - not like
ethnomethodology or activity theory (etc., etc.)
2. information focused: important in designing technology
focus of distributed cognition
 use of the term
 Generally used to mean approaches to the study of the division
of labour over multiple resources:
Other
people, artefacts and/or situation
 it’s about cognition
 information representation and processing
 …but with a difference in the boundary of analysis
… …and consequently, data collection methods applied
 Humans
as skilled co-ordinators of representations
(internal and external)
 and not necessarily directly performing mental processing on
that information
i.e.
applying general-purpose (co-ordination) skills and not case
by case planning/ execution
representation and the IP metaphor
 cog
sci: human mind operates through computational
mechanisms
 problem solving => changing problem representation changes the problem
 successive transformations on a representation can transform initial state
into desired state
 involves a computational transformation:
 of a problem state (i.e. representation of the problem)
 from start state through a ‘problem space’ (with resources and constraints)
 into a goal state
 through propagation of representations across various representational
structures
 cognitive architecture
 in human minds, representational structures are neural pathways
 DCog: no distinction between representational media – internal or external
– forming a system ‘boundary’
how is it cognitive/information processing?
extends study of individual
cognition, looking at
‘internal’ cognitive
representations and
processes…
…to the study of people,
their organisation and use
of external resources as
representations and
processes
Representations
INPUT
Process
INPUT
Processes
Representations
OUTPUT
OUT PUT
from Halverson, 1995
the ‘problem’ with traditional cognitive
science
 posits
we can understand human action without
reference to the outside world
 we just need to understand the mind, then we can understand
interactions (physical and social) with the world
 there
 and
are some problems with this(!)
what if we don’t yet fully understand the mind? how
can we do anything useful now?
Socially Distributed Cognition: the
social organisation of group problem solving
 group
activity is a computation realised through the creation,
transformation and propagation of representational states
 SDCog allows us to examine how such computations are organised, and
how representations act as intermediaries in collaboration
 many
ways to organise the system to distribute the computational
load
 some better than others (speed, processing resources required, proneness
to error)
 division of labour
 determines the computational architecture of the problem solving unit
 establishes the resources and processes brought to bear on problem
representations
 system can do adaptive structuring
 organising and reorganising physical and cognitive artefacts in the
environment
 modifying the social context
‘doing’ DCog
 unit
of analysis: the ‘functional system’
 individuals, cognitive artefacts/media - and their relations
 boundaries set by analyst (! - note theoretical implication - !)
 examines
‘information-representation’ transitions
 equivalent to examining a systems’ mental state
 many
approaches to ‘doing a DCog’
 a framework not a method
 generally, data collection is observational - the ‘cognitive
ethnography’
 Example:
 ConsCo
example 1 - media transformations
Ke y
Expected measurement
Height
Actual measurement
Length
shows key representational
transformations and
co-ordinations in the computation in
resolving the problem and
communicating information to the
senior engineer
Site
information
Table of s ite
measurements
Section view
sketch
Geotechnical
readings
Original plan
view drawing
time
Table of
measurements
from drawing
Post-it note
explanation of
data
Annotated
sketch to senior
engineer
example 2 - social co-ordination and cross
media transformations
Senior engineer (SE): ‘If you look here, there’s a barrel run there’
<points at sketch generated in the meeting of a section view through a
design structure>
Temporary works design co-ordinator (TWC): ‘Yes I see’.
SE: ‘So if we dig here...’ <he holds one hand to the sketch and runs a
finger on the other hand along a permanent works drawing (in plan
view) beside the sketch, indicating a line of reference>
TWC: ‘No you can’t do that because of drainage problems...’ <pauses>
‘...No, no, I see now’.
SE: ‘So if we cap these piles here...’ <indicates several points on the
sketch>
TWC: ‘Yeah. OK. Lets do that’.

common understanding: cross-referencing external representations
 bringing together and aligning representations collaboratively
 co-ordination of representations is mediated by the senior engineer using hands
to demonstrate relationship between drawing and sketch

allows him to indicate where the digging (on the sketch, seen from the side) would
have to be performed on the site (on the drawing, from an aerial view)
 physically using his body to mediate this, he creates a new, shared viewpoint of
the information on the two media
Informational transformations
Verbally encoded
mental representation:
“ …there’ s a barrel
run…”
Co-ordinates
representations
by pointing at
section view
sketch
Newly
generated
section view
sketch
Plan view
drawing
Co-ordinates
representations
(physical and
mental): “So if w e
dig here...’, crossreferencing drawing
and sketch with his
hands
(Future action)
generate design
brief to send to
design engineers
off-site
some recent work in SDCog
space as a computational resource
(spinelli et al., perry et al.)
‘more than just another factor influencing behaviour
in a setting’…and worthy of investigation in its own
right
embodiment is spatial!
 DCog
examines the role of artefacts, or ‘objects of coordination’
 but these do not just exist socially…
 but in real places, that carry meanings, and provide spatial
constraints that afford particular forms of use and
interpretations of their meaning

 how?
previous work by Kirsh, but supporting individual action
By physically:
1. orienting problem solvers to information (awareness)
2. structuring social organisation (cognitive architecture)
3. structuring collaborative computations (co-ordinating
representational transformation)
1. physically orienting problem solvers to
information

artefacts in line-of-sight support deixis: providing shared
reference in social interaction
a)
‘angular visibility’

b)
making visible the computational structure

c)
artefacts can be oriented to show to only those people within visual range
as a filtering device (i.e. impacting on the cognitive architecture)
orientation of other people to artefacts can provide actors with cues to
interpret the division of labour (i.e. the computational structure of the
distributed cognitive system) - ask: who is directing their gaze to what?
physical foregrounding and backgrounding: ‘zooming’

space is limited: moving artefacts sets visual limits on the number of
‘chunks’ of material in close proximity for detailed examination and
discussion
2. physically structuring social organisation

using space to structure the division of cognitive labour by
managing access to information artefacts
a)
physical division of labour

b)
moving artefacts allows them to be allocated to people/sub-groups
 a physical hand-off of responsibility for work (e.g. removal from a
common to a personal space)
simplifying co-ordination of parallel collaborative work
processes

by manipulating proximity and access to artefacts, information is no
longer available for processing by others
 flexible and contingent distribution of task responsibility without
requiring hierarchical protocols to be imposed or negotiated
3. physically structuring (collaborative)
computations

spatial arrangement of information artefacts is important in the
performance of the computational activity
a)
a unique physical reality (physical artefacts can only be in one
place at a time)

b)
spatial placement can constrain the order of action

c)
reification of spatially-related abstract principles (e.g. useful in
categorisation, or ordering)
ensuring mutual intelligibility (e.g. one thing on top of another, on top of
another)
mapping spatial structure to symbolic structure

structural arrangements of physical media tends to map onto conceptual
arrangements of information (e.g. top to bottom, left to right, close to far)
returning to broader issues
application in design,
limitations,
current status and
conclusions
application in HCI and CSCW design
 makes
visible the mechanisms co-ordinating representational
transformations
 insights offered into how cognition is & can be distributed across people
and the (increasingly) smart environments that they work and play in
a
description of the informational characteristics of work
 shows representational properties and functions of media
 important in developing and introducing I.S.
 highlights information bottlenecks and communications breakdowns
 and
where not to introduce digital technologies
 cannot be applied directly to HCI and CSCW design
 creative interpretation is necessary
 describes work in informational and computational terms
 gives system designers a stronger model of work
 frames social and organisational p-s in terms of representations and
processes: terms well understood by systems designers
limitations and applicability
 developing framework (mid-1990s)
 not an established discipline
 weakly-defined set of acknowledged characteristics and boundaries
 painstakingly
slow, needs expert knowledge
 DCog appropriate for analysing problem solving
 but not all situations are best described as p-s
 no
clearly framed way to do a DCog analysis
 reasonably robust theoretical framework,
 but not prescriptive in its application
 …nor how to apply it to systems design
 how
is this co-ordination achieved by agents/actors? (Latour)
 is DCog an emergent property of activity systems?
… or just a useful analytical device?
… … but is this any different from criticisms of GOF cog sci
DCog in broader use
date
of primary reference: 1995 - not superseded
widely
cited in HCI literature, rarely used in anger
(though freq. cited)
DCog paper: Hollan, Hutchins and Kirsh rated #10 ACM
download of the month (Oct 2006)
and most is theory-based
less
commonly discussed within cog sci community
although not considered too controversial any more
clearly
still relevant, but issues...
conclusion
emerging theoretical framework
 goes beyond the individual - functional system is an
information processing unit
 tries to take account of work context
 identifies co-ordination and breakdowns

 identify some areas that have not been explored
deeply (space - the final frontier?)