A Case Study of User Profiling in Living Memory

wackybabiesΛογισμικό & κατασκευή λογ/κού

14 Δεκ 2013 (πριν από 3 χρόνια και 9 μήνες)

97 εμφανίσεις

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


1

A Case Study of User Profiling in Living Memory




Living Memory Deliverable 2.5




Authors:
Eric Laurier,

Angus Whyte,


Kathy Buckner,


Tom Shearer,

Katie Bates


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


2




Introduction
.



During the three years that QMUC have been working on the Living Memory
project, our
task has been to contribute an understanding of social and cultural practices in a suburban
neighbourhood
1

to the design of prototypes intended to support interaction across that
neighbourhood. Our work has combined ethnography with action res
earch and
participatory design, in common with many CSCW
workplace

studies. However it differs
in important respects from these. We will describe three major differences in this case
study report, and how we have tackled them, with examples of what we ter
m media
archetypes and agent interactivities. An outline is given of the hybrids of ethnography
and action research we used to articulate aspects of current practices and relate these to
the evolving design proposals for a ‘Living Memory System’. In the ce
ntral section of
this report we pick up on the potential of these hybrids for informing the design of
user
profiles
for multi
-
site public networks.


A full description of the Living Memory (LiMe) system architecture is beyond the scope
of this report, wh
ich, as noted above, focuses mainly on user profiling. Nevertheless
giving an account of the LiMe project aims and design approach was just as integral to
our research on user profiles as it is to our discussion of it here.


Living Memory aims to provide
the members of a given community who live and work in
a particular locality with a means to capture, share and explore their collective memory,
with the aim to preserve and interpret the richness of local culture
.
2



Addressing this aim involved five geogr
aphically separated (Edinburgh, London,
Eindhoven, Brighton, Milan and Paris)
3

research partners in these four tasks:
-


1.

Identifying the needs of a local community;




1

At the outset of the project the term 'community' was used as an analytical term in our research to refer to
a city neighbourhood whilst also blurring this sense of community with that of connected communities
which has a specific meaning for sof
tware agent design and community
-
ware which was a new variant on
groupware aimed at serving teams of users. Given the object of our study we switched its label to
neighbourhood which was both less confusing given the many sense of community being used and
also less
value laden (since community normally implies 'good' community). Nevertheless we were still interested in
communities (plural) as constituted by socio
-
material practices, what Wenger [Wenger, 1998 #256] calls
'communities of practice'.

2

This st
atement serves the intriguing reflexive function of being the sentence which Living Memory
members have used as the means to capture, share and explore their collaborative research by reference to
this 'credo'. We used it in the initial funding application
, in most of our academic publications and
presentations, in our publicity brochures that we produced for distributing in our study area, and in giving
face
-
to
-
face explanations during our fieldwork. It is the 'text' which have struggled hard to translate
into
various other more detailed texts (i.e. specification documents, project reports etc.) and also into other
things such as coffee tables, software agents, postcards and websites.

3

One of the partners was based in 2 sites
-

Eindhoven and Brighton

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


3


2.

Designing interfaces to support content management and tools for intelligent memory
manage
ment;


3.

Demonstrating the relevance of a "living memory";


4.

Drawing methodological lessons which may be applied in developing products or
services in this domain.


It is worth bearing in mind that the project partners come variously from backgrounds in
soci
al informatics, ethnography, human
-
computer interaction, industrial design (as craft
and commercial activity), and the engineering of software agent technologies. Our focus
in this paper is on the first and last of the above tasks, and on displaying how we

have
tackled the relation between ethnography and design.


As described above, the project 'aims' beg certain questions: for example about what is
meant by the terms "a given community", "collective memory", and "local culture". In
this report we do not i
ntend to present a formalised theory of these terms, but we outline
below how such theoretically
-
informed notions helped to articulate a design agenda in
collaboration with our project partners.



Concepts of memory, culture and community


LiMe’s notion
of memory borrows from connectionism (e.g. Edelman, 1987, Cilliers,
1998) in that a "living memory" is envisaged as a distributed set of computer mediated
activities, the threads of which may be recalled (see Fig.s 18
-
20), and the associated
content re
-
pre
sented, depending on where people are and what they indicate an interest
in. This could be described as a form of
social navigation

(Munro et al), mediated by
interfaces that are integrated into the places where people meet to pursue their interests


situ
ated social browsing
. The cultural activities we aim to augment, and the memories we
are concerned with re
-
calling, are primarily those which are
informal
, make no claim to
be objective, and are mainly related to user’s
voluntaristic

pursuits rather than
t
o work
patterns and decision making

in organisations. Our emphasis is also on augmenting ‘real
-
life’ links in physical places, as opposed to virtual ones, though as we shall see later 'real
-
life' and 'virtual life' are utterly entangled in practice. Our ge
ographical focus is on a
suburban

area of Edinburgh, Scotland, centred on the historical village of Corstorphine.
By a "given community" we refer to this area, one that has historical and current
connotations of neighbourhood for our informants.


The pro
ject’s concern with the sharing of knowledge and experience between people in
this neighbourhood has obvious similarities with
community networking

(Schuler). The
settings we have focused on;
the local public library, the home, schools and a shopping
mall
,

are those which are typically sites of public and private access via personal
computers (PCs) to civic and commercial networks (i.e. CapInfo the city council's
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


4

information service, the world
-
wide web, e
-
mail accounts etc). In contrast to these
terminal ge
ographies LiMe put more emphasis on the prototyping of novel forms of
interaction that promote and enhance distributed and multi
-
artefactual links between
ordinary people, rather than on, for instance, implementing current web tools on pre
-
existing termina
ls.


The design of software, networked artefacts (i.e. electronic tokens (see fig.1)) for
distributed applications, to be used by socially mixed neighbourhood residents in diverse
settings, requires a high degree of flexibility to be built in to the archi
tecture that inter
-
relates human and software roles, procedures, social groups, and content creation. The
notions of software agents and allied perspectives like component
-
based design are
commonly seen as potential means to provide this flexibility (REF).







Fig. 1 From 'living interfaces, interacting with the LIME system' (Doc No.
LIME/908020/99002/LL)


The LiMe project design agenda included the implementation of broadly characterised
roles that are commonly seen (in the domain of software engineerin
g) as appropriate for
delegation to ‘autonomous’ software agents. These include
matchmaking

users with
content that relates to a profile of their interests, and with other users with similar
profiles.


QMUC's focus in the project is on articulating curre
nt practices and assessing how these
fit with the design agenda, rather than on formulating design proposals or a theory of
‘connected community’ (
but see

LiMe Deliverables 4.2
-
4.5 & Rutgers 1999).


Ethnography is primarily concerned with "what is", descri
bing current practices and
technologies, rather than with "what might be" if these were redesigned. Not surprisingly
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


5

the relation between ethnography and design can be problematic (Plowman et al 1995,
Grudin and Grinter 1995, [Shapiro, 1994 #295]), as the
re is nothing inherent in
ethnographic "rich description" of current practices that makes the design implications of
changing community activities anything more than informed speculation. Moving
beyond informed speculation commonly involves a hybrid of et
hnographic description
with participatory methods that involve prospective users in evaluating the relevance of
design
-
led changes (e.g. Greenbaum and Kyng 1990). The next section describes how we
have sought to inform design through just such a hybrid. W
hy our hybrid has taken the
form it has should become clear by pointing out some key differences between workplace
settings, the usual the focus of published studies, and the settings for our current research,
differences that raise critical issues for the

ethnography
-
design relationship.



Workplace Studies in Public Settings


Three important distinctions can be made between workplace studies and the "user
community" we are concerned with; the scope and range of settings, the focus on
informal communicatio
n practices, and an explicit ‘creative direction’ role in developing
concepts that are a radical departure from current user practices. Below we summarise the
issues these differences have given rise to.



The range of user settings


Our research setting
is a residential suburb with a population of around 20,000 people,
not including those who work in it or pass through it. The need to select from the range of
public settings that we could investigate was apparent very early in our work. Our
selection of a

library, school, pub and shopping mall could be made on the grounds of
their local significance and the potential for translating our findings to other
neighbourhoods and also as a response to the conceptual initiatives of WP3 (see LiMe
Deliverable 2.1,
Section 2). Selecting
specific practices

to study was far more
problematic. By comparison, workplace studies predominately feature relatively clearly
bounded locations, prospective users, and practices.




Informal practices


The focus of our ethnographic

studies is on "informal communication" practices, rather
than formalised work procedures (though not to the exclusion of them
4
). In comparison
workplace studies normally have the opposite emphasis, even if they commonly treat the
relationship between docu
mented procedures and the socially
-
constructed reality of work
as an open question (e.g. Suchman, 1987). From our perspective, what is "informal" and
what is "formal" is also an open question. Notions of practice normally entail some



4

Par
ticularly so in relation to the more formalised aspects of voluntary work, the running of local charities,
the library, the school and so on.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


6

concept of ‘rules of
engagement’ that can be found in particular interactions. The point
here is that in public settings like libraries, bars or shopping malls the rules of
engagement are rarely codified, except for those that work there. With the exception of
local structures

of public administration (not a focus of our studies), in an open
community there is no equivalent to the institutional texts that, even if they are only
‘useful fictions’, serve to characterise work and the terms of organisational discourse as
at least a

starting point for the redesign of workplace information systems.


This has presented significant problems of
access
, in two senses. Firstly, although
workplace studies can be negotiated through the managerial structures of the participating
organisation,

providing access to observe what people do, and record what they say about
what they do, this is inherently more difficult to negotiate when many sites and
organisations are involved. Secondly, the nature and content of interactions in public
settings is
(paradoxically) more difficult to access, because these interactions are often
fleeting, many may not be repeated on a regular basis and their content is private.



A radical design agenda


Thirdly, our research has to integrate an understanding of curren
t practices with attempts
to change them in ways that are
relevant

to potential users. This is what participatory
design studies usually seek to do, and it is not unusual for design research to focus on
radical departures from users’ current practice. It i
s less usual to attempt both and, in
practice, we have found that mediating between a "top
-
down" design agenda and
"bottom
-
up" enquiry into user practices is a significant challenge. The rationale for
having a top
-
down design agenda is that radical departu
res from current practice are
rarely requested by users, and they also rarely emerge from ethnographic studies which
are often seen as conservative in their design implications [Shapiro, 1994 #295]. The
"ethnographers dilemma" is that the more fine
-
grained

the understanding of current
practices, the more that ethnographers are likely to see any technological intervention as
disruptive (Grudin). With a "radical remit", one that aims for technology that is both
relevant to our prospective users and a step
-
cha
nge from what they currently use, design
concepts have to be envisioned (ref


campiello). The challenge has been to make this
top
-
down design agenda textually explicit, let alone manifest in artefacts that connect
with life as it is lived in the neighbour
hood of Corstorphine.



An Overview Of Our Research Approach


Our general approach is informed mainly by Suchman and Trigg (1991), which in turn
acknowledges the complementary roles of action research and ethnographic enquiry
developed in "Scandinavian"

approaches to participatory design. Its general form is
illustrated in Figure 2 below.



LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


7














Figure 2. User studies from three perspectives (adapted from Suchman and Trigg 1991)


The figure depicts the perspectives of ethnographic description, i
ntervention and
evaluation between which some of our multiple roles in the LiMe project can be placed.
Most of our work involves some element of each with varying degrees of tension to fulfil
the requirements of all. We outline below what this has meant in

terms of each
perspective.



Stereotypes and ethnographic description


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


8

In the previous section ethnography has been talked of as if it were one unified
perspective, when we should point out the particular forms that we have adopted. Early
in the project
our concern was to draw broad
-
brush profiles of each setting (bar, library,
shopping mall and school) and of the neighbourhood’s general demographic and physical
characteristics. When our conceptual design partners produced scenarios, envisioning
how a ne
ar ubiquitous
intranet
-
like

system might be used in the future in our selected
settings, we embarked on "mini
-
ethnographies" of each setting in six
-
week cycles. Using
semi
-
structured interviews we identified narrative themes from our informants'
reflection
s on their everyday concerns, their orientation to other residents in the area, to
community groups, and to shared places, events and media (see Appendix 1).


This form of qualitative research, involving the ‘coding’ or categorisation of themes in
intervi
ew transcripts, is commonly associated with a
grounded theory

approach (Glaser
and Strauss, 1967). This would be appropriate if our aim were to generate a substantive
theory of interaction in this neighbourhood. Our aims however were more modest
; to
provi
de an outline understanding prior to design work, a ‘reality check’ on the
plausibility of our partners design concepts
. These roles are consistent with what Hughes
et al (1994, 1995) term "quick and dirty ethnography" and "evaluative studies"
respectively
. Our over
-
riding objective was an empirically
-
based and theoretically
-
informed evaluation of how (and whether) a prototype "Living Memory System" fits
within its intended settings.


The drawback of our early mini
-
ethnographies was that they were, in effec
t,
stereotypical

profiles of the settings and the people who used and produced them (though we will re
-
examine place
-
based profiles later). As such they were not detailed enough for our own
purposes of ‘thick description’ [Geertz, 1983 #406]. They did prov
ide broad portraits,
composed from interview data, of our informants demographic characteristics, of what
they said they did, who with, and where (see Appendix 1), though as we shall see later
these portraits were still overly narrow to be of practical val
ue for user profiling!


Stereotypical though our views of the settings were, they allowed us and our design
partners to concentrate on more specific settings, prospective users and practices:
-



1

Publishing activities of children (aged 9 to 12) in and from

primary schools;


2

Pubs and cafes as settings for informal interactions in public space;


3

Home
-
based exchanges and encounters of locally resident seniors and their
orientation to voluntary groups.
5


Stereotypical profiles provided a sounding board for Dom
us Academy's design concepts
(see LiMe Deliverable 3.2?), but were not rich enough for interaction design purposes.
They gave little insight into what people do, or specifically how Domus Academy's



5

A more detailed report on our research in these settings can be found in LiMe Deliverable 2.4.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


9

'design scenarios', Imperical College's 'Multi
-
agent syst
em', Philips Design's 'connective
tissue' (Rutgers 1999) or 'living interfaces (Matthews 1999) might be brought to fruition.
6



Our approach to seeking thicker descriptions of practice in the settings mentioned above
was
ethnomethodologically
-
informed
. Eth
nomethodology as an investigative exercise
builds a highly detailed exposition of the methods used by ordinary (and expert) members
of society to make sense of, account for, and reproduce the particular circumstances that
they find themselves in (e.g. Garf
inkel, 1967, [Wieder, 1974 #298]). It makes very few
a
priori

assumptions, is not involved in theory
-
building in contrast to most other
professional social science and is rooted in the lived accountability of any social
situation. To complement our 'low t
heory' approach, our descriptions have also been
informed by work on the socio
-
materially distributed nature of agency (Latour, 1997)
(sometimes called'actor
-
network theory'), the communities
-
of
-
practice perspective
(Wenger, 1998), and concepts that have e
merged from social informatics, notably
boundary objects (Star, 1989) and genre repertoire (Orlikowski and Yates, 1994).


Many of the practices we observe in everyday settings are mundane [Pollner, 1987 #419],
in the sense that they are
already familiar

t
o us as residents of broadly similar
neighbourhoods, in some senses they are taken
-
for
-
granted and above all they are
'worldly' in scope and orientation. The forms of investigation mentioned above are useful
for sensitising researchers to just these oft' o
verlooked everyday and yet crucial aspects of
life in a fairly unexceptional suburban neighbourhood. They help to untangle just which
of the tacit assumptions and routine interactions involved would be necessary to ensure
the viability of the project proto
types (i.e. social orderings a prototype can depend on),
and which might be changed by them (i.e. social orderings a prototype will disrupt or
otherwise mangle).


Our professional academic resources do not however overcome the issues of access
mentioned
earlier, nor should we expect them too.
7

Accessing communities of practices
involved to put it bluntly: joining in with what the locals do and learning about how they
do it. Just like any other neighbourhood resident we participated in only a selection of
the
possible activities going on in the suburb and indeed probably ended up being much more
highly involved than the average suburbanite.



Scenarios and prototype evaluation


Given the issues mentioned in the introduction, involving users as design parti
cipants
presents several difficulties, exacerbated in our case by the fact that our design partners
are located hundreds of miles from the selected user community. Although it would be



6

From the number

and variety of speculative designs of Living Memory in the pre
-
prototyping phase being
produced by each partner, a sense of the difficulty of QMUC's and Philip's Human Factor's work in
communicating what defines a Living Memory system to residents, who co
uld see little relevance to their
current activities, can be appreciated, a point which we will return to later.

7

If someone calls at your door and tells you they are 'ethnomethodologically
-
informed' and believe in the
'power of mundane reason' would you

let them in?

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


10

desirable to involve users directly in designing their profiles, testi
ng them and ultimately
sharing in the benefits of the development of Living Memory, the research as a 'long
term' and more 'visionary' European project has inevitably settled for "proof of concept",
without long term trials of user
-
profiling nor full imple
mentation of the poly
-
conceptual
system.


Securing the commitment of a range of members from the many settings implicated in
design concepts, in continuous re
-
design of prototypes was unsurprisingly unfeasible.
Potential users have nevertheless been contin
uously involved, but mainly as ethnographic
informants (an entity quite different from a 'user'), providing their views on futuristic
LiMe scenarios and working prototypes (see LiMe Doc XXX 'LP3 Testing' and 'LP4 in
the field').


Scenarios
, produced by Dom
us Academy, have taken the form of visually rich
storyboards and very short narratives. Firstly, storyboards were developed for the library,
public library, public bar and shopping mall settings (see LiMe D3.2/3?). These depicted
the use of large interacti
ve screens by site users, and smaller consoles with which the
human agents in charge of each setting could guide and control the content available to
members of the public. In each there was a clear relation with a range of activities which
were typical of

the setting, and which drew on connections with other settings. For
example in the school setting, schoolchildren’s activities in compiling articles for a
school gazette were shown, augmented by mobile devices for gathering and transporting
their material
s, and for relating these to relevant content from both local and broadcast
sources. In addition, roles for software agents in disseminating content, according to the
preferences and routines of users distributed throughout the neighbourhood, were
articula
ted (Pitt, Mamdani , 1998).


The scenarios were evaluated in collaboration with our design partners, by users of each
setting who were attending a local fair, and through interviews with local residents
identified through mailshots and by telephone. In eac
h case we identified ‘pros and cons’
of the futures envisaged, with the short narratives being used to refine design priorities, in
a similar way to the early stages of the task
-
artifact cycle (Carrol et al...)


To involve prospective users more directly i
n design we have also experimented with
co
-
design sessions
, in collaboration with our design partners. Here we recruited active
members of various local interest groups to map out the communication tasks they
associated with organising a campaign to safegu
ard a beauty spot against property
development, and with publicity for a fundraising fair. Using the "interface metaphors
game" described by Wildman et al (1993) for interface design, we explored the
mappings between these tasks and those that our partic
ipants associated with more
‘everyday’ activities such as ‘working at a hotel reception desk’. This was seen as a
useful way of identifying ‘clusters’ of communication tasks, such as ‘brokering skills and
available people’ that would need to be accommodate
d in the system interface and its
underlying architecture.


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


11

In years 2 & 3 of LiMe, 4 prototypes were developed in response to the feedback gained
from these evaluations. Again these have been evaluated with potential users as
informants (see LiMe Doc.
s XXX 'LP3 User Testing' & 'LP4 in the Field'). Sessions in
laboratory settings (constructed to look as little as possible like computer labs, see fig.3)
have allowed informants to explore the interfaces in some detail, and comment on their
usability. In t
hese sessions they have also discussed how the prototype, if it was
implemented pervasively throughout the neighbourhood, might relate to their current uses
of noticeboards and newsletters, and the editorial issues that their use of it could entail.






Fig. 3
Lab Testing a Prototype element of a Living Memory System (AKA 'a slice of
LiMe').




Testing the prototypes in field settings, with people who ordinarily use them and who
happen to be there, have told us more about how a better fit between prototy
pe and
setting may be achieved. In its most recent iteration, the prototype ‘living memory
system’ has been set up in a school, hotel bar, medical centre waiting room and library
(see fig. 4). In short sessions (typically several minutes) potential users

have commented
on interface usability, content relevance, physical aspects of the system and its likely
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


12

usage. The negotiations required in order to set
-
up prototypes in these settings have
themselves told us much about the work that would be required for

these forms of
prototypes to become a permanent fixture in them (LiMe Doc.s XXX & XXX).






Fig. 4
Living Memory Prototype 4 (electronic café table, LiMe database, filtering agents
and tokens) in local lending library
. On the right are two potential us
ers of the electronic
café table trying it out and giving their opinion. On the left are two researchers from
QMUC listening and videoing the session. Behind the prototype library business goes on
as usual.





LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


13




Fig. 5
Living Memory Prototype 4 being se
t
-
up in a café
-
bar. With practice and training
it takes only around ten minutes to wire it together and start it up.



Intervention through action research


Action research differs from ethnography in that the understanding it produces comes
from reflect
ing on the results of action initiated to change a situation, rather than from
observations based on a minor participatory role in that situation. The action is taken with
the active involvement of a setting’s members, for whom it must have some relevance
(Checkland 1999, Checkland and Scholes, Avison et al 1999).


Our particular orientation to action research is (as Figure 2 makes plain) drawn from
Suchman and Trigg’s work, and in turn influenced by Schon’s work on reflective practice
(Schon, 1976). This
advocates that practitioners articulate their reflections on how they
tacitly frame their action and use material and conceptual resources in that action. We
have used the vocabulary and semi
-
formal notation of Soft Systems Methodology
(Checkland and Holwe
ll, 1998), to communicate some of these reflections to systems
design partners.


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


14

This stream of work has been necessary to complement the description of current
practices with an analysis of the changes initiated by Living Memory’s appearance.
Although ou
r actual interventions fell short of the envisioned concepts for a Living
Memory system, they articulated aspects of what our design partners wanted to build, and
combined these aspects with features that we judged to be directly relevant to our
prospectiv
e users and implementable with low
-
cost, off
-
the
-
shelf technology.


To describe this work we use the terms "media archetypes" and "agent interactivity" (for
a longer analysis see Whyte et al. (forthcoming).


Media Archetypes


These use off
-
the
-
shelf tec
hnologies to implement new artifacts (newsletters,
postcards, web
-
based "gazettes") that are intended to be relevant to members in
diverse local settings (e.g. schools, pubs, libraries). Their design and production
are based on an interpretation of
current

practices
, abstracting out elements of
locally used genres and relating these to an explicit design agenda. They are a
kind of field experiment, to test the relevance of particular variations of content,
form and media to users in their various settings.

Media archetypes have similarities with the PICTIVE approach (REF), and
case
-
based prototypes (REF) in that they use relatively ‘low
-
tech’ technology to
mediate between characteristics of current genres and characteristics of more
radical design specifica
tions. Media archetypes should address the practices of
their prospective users and incorporate the materials used in them. They should
also be usable in 'real' settings (i.e, they travel to their user's site, rather than
having users travel to a 'lab' or

other protected environment). If this does prove
to be the case, they will generate contextualised content, contradictions, limits,
and possibilities for more innovative prototypes, and help to establish settings
for the evaluation of these. Media archety
pes are interventions in settings and as
such they need to evaluated in their own right. Their relevance can be evaluated
explicitly with their users, and inferred implicitly. Inferences can be made from
ethnographic observation of their usage, and reflect
ion by researchers (acting
from a design perspective) on the everyday knowledge required to produce
them, make them available, and account for their presence to ordinary members
of the settings they are available in.



Agent Interactivities


Identifying
boundaries between human and software roles and information
dissemination tasks is a complex design problem in
workplaces
, and a
significant challenge when, as we noted earlier, the software architecture is
meant to serve
informal

roles, groups, and proced
ures. These are relatively
transient, often inaccessible to ‘outsiders’, and difficult to characterise without
formalising, elaborating and heirachicalising them to a degree that would make
the results irrelevant or unacceptable to their ‘users’.

As part

of our descriptive work we have enquired into how potentially relevant
software
-
agent roles are accomplished by human agents in real
-
time, naturalistic
situations. This involves taking the kinds of "insider
-
outsider" role normally
associated with ethnogra
phic participant
-
observation. In addition though, our
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


15

own work in producing "media archetypes" involves us reflecting on our own
roles as human agents, brokering the communication needs of local individuals
and groups, and, the diverse design, engineering

and commercial interests of
our partners. We do this with the aim of assessing the fit between this agency
and the logic that would need to be engineered into a software agent to perform
‘the same’ roles. And what is perhaps most challenging in LiMe is th
e constant
shifting of agency between users, software agents, and institutions.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


16


User Profiling


In this section of our report we will tease out particular strands of Living Memory's
research on user profiling that were innovative and challenging for all
the research teams.

Unusual Users


The first thing to note is that for nearly all configurations of " the user" in connected
communities and software agent research (with some notable exceptions i.e. research on
games, VR & wear
-
ables) the assumption is t
hat we have a user in a
workplace

or
institutional

setting. In such a setting, users will have defined roles and tasks such as
secretaries, managers, clerks, accountants, academics etc. As part and parcel of the
setting the user organises their work primar
ily via a keyboard and screen interface.
8

To
use a musical metaphor, wherever we find the "the user" they are always part of an
orchestra and they are always playing piano. One important body of work that has sought
to examine users out of formal settings
are the studies of technology in domestic settings
(i.e. family viewing of television) (Moores 1990 [Morley, 1992 #258]). These studies
produce a dramatically different sense of who "the user" is, since someone who is by day
a manager becomes a mother of t
wo once she is home at night.


Caught between two socio
-
technical settings (domestic spaces and institutional spaces)
Living Memory's users are neither being treated as family members 'at home' nor
institutional members 'at work', they are, instead, the m
embers of public and often
informal sites within a neighbourhood. The majority of the places in the latter half of the
Living Memory project were 'passing places' (i.e. bus stops, cafes, waiting rooms, notice
-
boards) where people's configuration in those p
laces was correspondingly temporary. In
each of those places their interactions would not be with a keyboard/screen interface but
would instead involve the use of touch
-
screens, identity tokens (see fig.1 & fig.4),
telephones, cameras, postcards and so on.

The ramifications of configuring 'the user' of a
Living Memory system without their interaction being based around a screen/keyboard
nor in relation to formal organisations nor households are far
-
reaching, promising and
novel, and went beyond the previous

experience of QMUC's staff.



Site As User for Concept Design Scenarios


In the first year of the ethnographic research, the production of user profiles coalesced
around 'sites' as 'users' rather than individual humans as 'users'. To build profiles for
Do
mus Academy' 'scenarios' (LiMe Deliverable 2,2, 3.1 & 3.3) QMUC investigated
typical users of shopping malls who were given roles by their presence in those settings



8

What Mike Lynch [Lynch, 1994 #222] calls 'digitality' by reference not to binary storage of data but
rather through the practical accomplishment of tasks via the use of
fingers

(digits) on a keyboard and
mouse and the co
-
ordinated reading

of a display screen.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


17

(i.e. shop assistants, security guards, tellers, customers). For the mall the equivalent o
f the
‘Living Memory User’ was the ‘Shopping Mall Customer.’ This key entity for the mall
was constructed through the administration of customer surveys into 'demographics' such
as age ranges, income levels and so on. A Shopping Mall's managers need not an
d could
not know every detail of their customer's lives, they tend instead to create their profiles in
terms of their 'average customer', total quantities of customers passing through theirs
doors by hour of the day, days of the week, weeks of the months,
levels of car ownership
to plan for the provision of parking space etc. In other words their construction of their
'users' is based on the aggregations, summations, associations, periodisations that they
use to order the running of the shopping mall. Our p
rofiling was done in a similar way in
an attempt to meet the needs of Domus Academy's Concept Design work (see Box 1):


Box. 1
-

Excerpts from Deliverable 2.1






























Lengthy negotiations were carried out between QMUC and Domus Academy

to agree
upon user profiles suitable for transferring 'knowledge' between partners (LiMe D.2 page
13 onwards). After four iterations of these profiles a format was agreed upon that was
suitable for Domus Academy's concept design work (see Appendix 1 and L
iMe D2.3

(QMUC carried out) “Mini
-
ethnographies” to collect and condense data on the
four sites selected for WP3 scenar
io development: a local school, library, pub
and shopping centre. The aim has been to profile site users, their interactions,
site
-
related activities and use of communication media.



Our work on mini
-
ethnographies and
user profiles

related to the LiMe con
cept
scenarios has identified ‘typical users’ of the interaction sites concerned, and
provided condensed information on the activities which currently go on there. WP2
has also initiated ‘pre
-
design interventions’, where we have become directly involved
in

communication activities related to the scenarios.


The scenario related strand of WP2 work has suggested areas of overlap between
site activities, which could be augmented in ways which develop the ideas in the WP3
scenarios, for example:
-




Activities in
volving child
-
care in the school, library and mall;



Activities involving school student ‘reporters’, their peers and parents in the shopping
mall, local charitable groups and pub ‘regulars’ ;


Integration across the sites of informal feedback on local eve
nts and issues, and personal
trading or barter of unwanted goods and home
-
based services.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


18

p14). Though
this

form of user profiling was not as it turned out suitable for user
profiling for Imperial College in their work on software agents or Philips Design for
either working with 'living interfaces' or 'connective tissue'. We will not c
omment much
further on this form of user profiling since it was not utilised in any significant form
beyond the concept design process.




Profiles of users as
connected

entities for Imperial College Acquaintance Models




Fig 6. A view of a LiMe user's
node produced by Imperial College (from LiMe
Deliverable 4.2)


The second major type of user profile which QMUC co
-
operated in producing was the
Acquaintance Model Interface (AMI) and Acquaintance Model (AM) which was based
on initial work by QMUC on mappi
ng social networks through SOgrams (Kathy ref?).
Fig. 6 illustrates the section of the AMI that displays the AM from a user
-
centric
perspective (i.e. 'Oscar' the user is at the centre of the profile).
9

However as one might
expect with a piece of software a
t such an early stage the AMI was not easily transported
from Imperial College's equipment and systems of support. In keeping with our action
research approach, as outlined earlier, the AMI was instead translated
back

into a 'media



9

For instructions on how Imperial College went about creating an AM via their AMI in terms of whether
that 'node' was a 'person' or an 'event', and if the former then what their relation was with the person, and if
the latter than a
brief description of the event, see D4.2 p14.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


19

archetype' (since the AM
I was a modification of QMUC's SOgrams) which did not have
to rely on keyboard, mouse and screen devices. During extended interviews with
researchers from QMUC respondents would create Significant Other Diagrams (SO
-
grams) (see fig. 7).






Fig. 7
A Sig
nificant Other Gram hand
-
drawn by a potential LiMe user


We can see obvious relations between figure 6 and figure 7, and can see that constructing
an acquaintance model is certainly not an impossible task. However the electronic
version pictured in fig. 6
also had certain additional qualities in since it was not static as
the drawing in fig. 7 is. The electronic AM was able to be linked to other AMs by the
Living Memory system and its centre could be replaced by a node which was previously
on the periphery
(see fig. 8 & see also Philips translation in Fig 18
-
20). Interesting
though such an idea was, it was not something we were easily able to allow people to trial
with our media archetypes (fig. 7 & 9).



In our initial research using SO grams (see fig. 10)

our interviewees were asked to
comment on the nature of the ties that link them to valued others by answering the
following questions:



-

How often do you communicate along this link?


-

How do you communicate along this link (by phone? by e
-
mai
l? face


to face?)

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


20


-

Apart from messages, what other kinds of exchange are made along


this link (e.g. gifts, services, helping out)?


We interviewed twenty
-
six participants of different ages, ranging from senior (60+)
members of local inte
rest groups, to members of the local Boys Brigade (an organisation
for teenage boys). In an interview situation with the assistance of the interviewer and the
guidance provided in terms of the questions above our respondents found SO
-
grams easy
to draw, ea
sy to annotate and, at a rudimentary level, easy to use as a way of clustering or
classifying relationships.


At a later stage of our initial research on SOgrams we integrated them with a computer
-
based concept
-
mapping package. From the rough evaluation we

carried out on computer
-
based SO
-
grams, using members of the research team, we identified a paradox. It was
clear that unless attributes of relationships are specified and constrained, then receiving,
reading or sharing a SOgram may capture users in a qua
gmire; since the descriptions of
relationships and the multi
-
media items associated with them were frequently too rich
and detailed for rapid use by the person that produced or their acquaintances that they
might share their SOgrams with.





Fig. 8
The e
xtended network of the electronic acquaintance model allows another
person's acquaintance model to be inspected?
10


In a later research initiative (see D2.4 ‘Making Acquaintance with seniors’) we returned
to the task of trying to producing networked user pr
ofiles, though this time with



10

There were many discussions during the project over how such privacy issues should be managed which
were not settled for practical purposes, thus whether the model would allow itself to be 'crawled' around as

html documents can be 'crawled' on the world wide web was not resolved.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


21

‘constraint’ and some degree of standardisation. In this latter research we only used
seniors rather than a more mixed demographic from the local area. The motivation given
to seniors for producing simplified SOgrams was that
they would assist in planning a
website for them (as well simultaneously providing an initial acquaintance model for
Living Memory).




Fig. 9
SO
-
gram made by hand with template. Size and colour were used to grade the
significance of people nodes (red hig
hest). Colour of line was used to identify frequency
of contact. The mode of contact was indicated using different stickers, marked with a face
(face
-
to
-
face), T (telephone), E (e
-
mail) or F (fax). Lines were drawn around nodes to
group people into 'zones'
.


It is worth noting that despite our having apparently simplified the production of
SOgrams/AMs the seniors we involved in producing them, without a LiMe member and
at home, found it surprisingly hard to produce summarised diagrams of their interests and

activities, even when it would seemingly be immediately useful to them in planning a
website (see D2.4 'Making Acquaintances and Making Acquaintance Models with
Seniors'). One of the possible reasons for the
non
-
production

of acquaintance models
during ou
r seniors research, apart from a person not having any acquaintances, is that in
our earlier research using Significant Other diagrams, the production of diagrams was as
part and parcel of
doing an interview

with a professional researcher, wherein the dia
gram
was produced with the ongoing guidance and commentary of the interview situation.


Interestingly when the interviews went particularly well the participants, particularly the
older ones, were sometimes overwhelmed because they listed and attempted to

connect
too many acquaintances. When the latter group of seniors were asked to complete similar
diagrams in their
own time
, even with clearly written instructions, they seldom did so.
The missing element, we would argue, was
the interview as the occasion
for the activity

without which the diagrams alone lost much of their meaning. We achieved contrasting
results when working with schoolchildren and it may be that schoolchildren's more
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


22

limited numbers of acquaintances may provide one reason as to why they f
ound it easier
to produce diagrams, though it is more likely that
the classroom as the setting for the
activity

provided the impetus for the production of the diagram.
11




Fig. 10
Representation of a user
-
centric social network produced using Netmap®.



O
ne of the questions that we were not able to answer is whether a software programme
such as Inspiration could have configured the activity of users in a similar way to being
interviewed by shaping the unfolding production of an acquaintance model and retai
ning
a determinate sequence by reference to which a SOgram’s meaning could be partially
recovered. As we noted already, in our earlier research using Inspiration we found that
the richness of detail that could be attached to a SOgram actually made the mana
gement,
preparation and subsequent utilisation very time consuming, when the point of an AMI is
that it would be as easy to use and manage as an address book. Nevertheless it may be
that diagrams such as the one shown in fig. 10 which are produced using N
etmap and
built on data gathered from users about their activities could be used by both professional
social researchers and possible also as a visualisation tool for Living Memory network
managers.


Motivating users to
explicitly

provide acquaintance mode
ls (and other forms of detailed
data on who they knew locally and the nature of their relationship with them) is thus one
of the initial obstacles to gathering acquaintance models which would allow software
agents to make
inferences

on the basis of pre
-
exi
sting friendships and interests (see
below) who might be introduced to who. Thus during our home interviews with seniors



11

It should be noted that the school children’s production of AM appropriate representations was generated
from a
third

form of activity


which was the classroom
-
based drawing of ‘me
mory mats’ (see LiMe D2.4
‘Classroom Observations’ & ‘The Reporter’s Organiser’). In other words candidate AMs can be produced
from a variety of different activities.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


23

we were faced with variations on the questions: "why should I make the effort to draw
this network?", and by association, "what possibl
e benefit can I get out of this extra
representation work?" A further related difficulty is retaining a
detailed enough
record of
the situation from which it was produced to allow user to backcheck what the model is
claiming to be a summary of. The issue o
f the accountability of the AM element of a user
profile is one that we will now turn to in greater detail.


The implications of Living Memory’s potential capability for creating new links between
residents in a neighbourhood was first made unavoidably app
arent to us in our work with
local schools, since they were protected to a greater degree than adults in terms of their
contact with ‘strangers’.


In December 1998 discussions with the teacher and head teacher of School
-
1 (see D2.4
‘School Gazette Media Co
nstraints’), they raised concerns about our proposals to try out
a form of user
-
profile matching which would put primary school class members in touch
with local people who shared an interest in the topics of their school gazette articles.
These concerns w
ere high enough to put a stop to the exercise. This meant not only re
-
thinking the basis of our involvement but also having to tackle the accountability of a
Living Memory system to the communities of practice it was intended to support.
Education authorit
ies in the UK place restrictions on the sanctioning and supervision of
contacts between children and adults from outside the school, as is the case in Edinburgh.
We found that education authorities also restricted access to their networks to legitimate
use
rs of an educational intranet.


We had initially framed our involvement with community members in terms of a "push"
metaphor, i.e. just as web services may push content based on a user profile, we intended
to "push" contacts with local people on a similar
basis. The objections, which had not
been raised until the prospect became a reality, led us to refocus on people "already
known to the children or the school", i.e. parents and friends of the children, or education
workers (including freelance workers or
people employed by community groups) who
have City Council approval for working with children. Under the City Council
regulations on the recruitment of volunteers, even parents of children at schools may only
become involved in work that involves direct s
upervision of children if they have gone
through a
registration

process, which includes a check by the police for criminal
convictions. The seemingly benign ambition of Living Memory to assist people in a local
area in meeting one another face
-
to
-
face mean
t that it was no longer just e
-
mail and html
files that were travelling in and out of the school but that accompanying these tokens (see
fig.1) would be children and non
-
registered local residents would travel to meet one
another face
-
to
-
face.


The issues
raised by our work with the school on making links between people on the
basis of acquaintances models goes beyond the special protected status of children within
schools. In our research the question was continually raised by local people
about who
was to

be made accountable for the links
.


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


24

In Fig. 6 Patrick is Oscar's ‘
boss’,

which is a relationship that is explicitly
agreed upon

as
part of their work setting, and is publicly available, carries various kinds of legal
responsibility etc.
whereas
in fig. 9
, in the top right corner, ‘me’ has a group of friends.
Whilst being a ‘friend’ carries all kinds of responsibilities which are quite different from
being a ‘boss’ it is also not pinned down in the same contractual and legal manner to that
which being a ‘b
oss’ in a workplace is. Also there may be divergent perspectives on just
how important such a friendship is to each party and what obligations come with it.
Coming to agreement on the terms of a friendship occurs in a different manner to coming
to an agree
ment on the terms of ‘manager’ and ‘managed’, or ‘boss’ and ‘assistant’ etc.
The most common potential use of Sogram
-
like maps of who
-
knows
-
who is in workplace
settings to look for weak connections in an organisational structure where it is likely to be
ad
ministered by the personnel or human resources department. Just who could be held
accountable for checking the status of friendships in a neighbourhood is harder to say. In
a business setting a third party (i.e. the human resources dept. can be held accoun
table for
their administered network map of the company), arguably in a community it would be
the friends themselves that would do so. There is an important distinction to be made
though since it is
public knowledge

within a business company or insitution
as to who
works for who and in what respect, whereas friendships are not necessarily accessible to
others in the same respect. Also as we noted above AM models could be used by
software agents to encourage people who had never met face
-
to
-
face before to s
trike up
an acquaintanceship. Yet the question is by what right does a software agent pass around
friends and how is its instrumental economic game
-
like exchange of people to be
regulated (and perhaps humanised!)



Agents amending the user profile on the b
asis of their inferences


Also it has to be born in mind that the Living Memory system was being granted powers
to make
inferences

about the status and quality of friendships. An existing weakly
inferential system which we reflected upon to shed some light

on this matter was British
Telecom’s (BT) ‘Friends and Family’ scheme. This system offers a telephone number
-
specific discount to customer’s initally on the basis of their explicitly nominated friend’s
and family’s phone numbers. A customer chooses ten of

their telephone numbers to be
discounted usually on the basis of the numbers they call the most (BT if they have
previous call records will tell the customers what their top ten numbers are). Once the
scheme is underway then from time to time BT will send

a reminder to the customer
asking them if they want to update their list of ‘friends and family’. So it seems that here
we have an equivalent for the kind of AM that Living Memory might implement since it
is a defined list of a customer’s (or user’s) frie
nds and family.


Yet the important distinction between BT’s billing discount scheme and Living
Memory’s AM is that BT’s scheme does not 'matter' in any significant way to a
friendship. It does not matter in the sense that I may or may not be on a friend's
BT list,
and


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


25

a)

I will probably not know this since I cannot under normal circumstances inspect my
friend’s telephone bills

b)

if I did know that I was not on their family & friends discount scheme I would not
assume that this meant I was
not

their friend since

I know it is only a crude measure
of the most expensive telephone numbers. On such a basis the dial
-
a
-
weather
forecast, internet service provider and so on may be on the list of friends and family.


Though even with BT's phone bills it may matter to a sha
red household (e.g. a family
home) where the most used numbers are highlighted and become the source of and cited
during arguments over who certain people are that they are called so frequently (i.e. a
parent suggesting that their child go round to their n
earby close friend's house rather than
calling them by phone). To take a more extreme example: if the Living Memory system
monitors e
-
mail or telephone exchanges between community members and makes
inferences on the basis of content and frequency of contac
t between those members about
who is friends with who, and two of those members are having a secret affair (i.e.
involving infidelity with their partners), then we can guess that making these interactions
‘visible’ may cause all kinds of problems.


To sum
marise then, user
-
profiling involving acquaintance models though undoubtedly
innovative raises 3 significant issues for community systems:


1.

How are mutual agreements reached on what the nature of a relationship is between 2
residents.

2.

Who is able to inspe
ct or otherwise utilise the interconnectedness of the web of
acquaintances created by the Living Memory system.

3.

How are the
inferences

about who knows who and in what capacity made accountable
and agreed upon by the Living Memory system
for

its users (as w
ell as for its own
purposes).




Profiles of user interests for Imperial College Software Agents




LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


26



Fig. 11.
Identity of user for the purposes of LiMe system.




Fig. 12
Interests as listed for the LiMe system.


The third form of user profiling (see fi
g.s 11 & 12) which QMUC investigated
ethnographically is perhaps one of the most commonly used: lists of interests. As we can
see in fig. 12 these interests consist of single words typed into the system, with a date
when they were entered and corresponding

date when they will expire. Such lists were to
be used by the Living Memory system to answer potential queries from a local resident of
the kind ‘is there anyone in the neighbourhood who is interested in travelling and
sailing?’. The Living Memory system
would not only act in response to queries from
local residents but ‘matchmaking’ software agents would pro
-
actively match up users
with similar interests and suggest that they get together.


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


27


Box 2


Excerpt from LiMe D4.4 'The Role
-
based specification of

a PSA'
-

describing the
sub
-
roles which are components which allow a personal service agent to play.


In terms of QMUC's ethnographic work it can be seen that our initial profiles of users
(see Appendix 1) generated out of questionaire
-
based interviewing

with Corstorphine
residents were too detailed for the purposes of user
-
profiling for Imperial College's
'Multi
-
Agent System' prototype. Indeed although the filled
-
in questionnaires are
documents out of which a human reader could build a profile of a user,

their more loosely
formed structure would undoubtedly pose great difficulties for a machine reader to build
a profile.


Greatly simplified user profiles were used during the testing of Living Memory Prototype
4 (see LiMe Doc. XXX). These consisted of keyw
ords which users cut and pasted from
pre
-
set comprehensive list into their own short 'personal' lists. The comprehensive list
was generated from an initial search of the 'content' in the prototype database by ??? (was
it an agent or some other indexing pr
ogramme). Unfortunately due to the restricted form
of the content in the prototype (i.e. a small number of items, each of a very short word
count) and the novel way in which the content was displayed on the screen we were
unable to learn a great deal about

how users might manipulate their profiles.


Once again QMUC's main method for investigating the implications of user
-
profiling that
allowed for the matchmaking of interests in a neighbourhood was through action research
and media archetypes. To this end w
e gathered lists of interests of local residents via
interviewing at home and did our own matchmaking exercise between relative strangers.
This exercise is more fully reported in Deliverable 2.4 'Making Acquaintances and
Making Acquaintance Models with Sen
iors'. In summary we brought together a married
couple and another man from the Corstorphine area who had not met before for a tutorial
in publishing on to the world wide web (fig 13). Before the tutorial began we introduced
the couple to the man and poin
ted out that they shared an interest in amateur radio and
playing music. We were in effect thus performing the 'matchmaking' role which one or
more personal service (software) agents would do as part of the Living Memory System.



3.1.5 Mat
chmaker

The tasks in the matchmaker role support the agent to access the acquaintance
model and the profile of the user in order to match the user with other agents and
users. These may involve the interests of the user, for example the user might ask
whe
ther there is anybody else in the system that is interested in gardening. When
such a request is made, the agent contacts other agents belonging to the user’s
acquaintances, in order to find possible matches. If no match is found at that level,
the agent i
n that role tries to ask agents that belong to acquaintances of
acquaintances, and so on. The results of this process are reported to the user.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


28




Fig. 13.
Matchmaking

LiMe users with shared interests in their profiles
-

the couple sit
opposite with the husband's hand visible to the right and man sits closest to the camera.


We sat the man and the couple around a table, as show in fig. 13 for an hour before we
began the

actual tutorial. This gave them plenty of time to 'make one another's
acquaintance'. As we have described in our full report this involved the man and the
couple showing not just that they shared interest but how despite sharing interests they
diverged in

their involvement in these interests. What was surprising to us was in fact
how much they appeared to disagree about various common topics. However this was in
some ways a reflection on our having introduced them to one another as 'people with
something
in common' which was the occasion for the discussion of these topics. We
were aware that by the end of the tutorial, the couple and the man had managed to avoid
showing the basis for a friendship, and had indeed very skilfully avoided making friends
with
one another without being rude or causing each other substantial embarrassment.


Part of what 'matchmaking' on the basis of user profiles does, to an encounter between
unacquainted Living Memory users, is make the nature and likelihood of these peoples'
fr
iendship a
primary framework

[Goffman, 1974 #433] for the encounter that follows. It
has changed the emphasis from a situation such as we might find at a horse
-
riding club
where the primary framework is still horse
-
riding and new acquaintances may or may n
ot
emerge from encounters while at the club, to one where people pursue acquaintanceship
first before pursuing their interests.
12

Putting the cart before the horse! We had made
(and similarly LiMe would make) the situation problematic by implying that the
couple
and man
should

become friends because they had things in common according to their
listed
one word

interests
13
. Fortunately our internet tutees were highly accomplished at



12

It is worth noting that historical function of city clubs (i.e. the Groucho in London)
in the UK has been
about gathering people with common interests (and backgrounds) together but in a place and without
necessarily specifically introducing one member to another. Also these clubs make withdrawal from an
introduction possible, do not tend to

draw on one neighbourhood but from a more dispersed area in the city
and finally often work by excluding certain people.

13

By the very formaluation of an 'interest' as one word, rather than as a phrase, or a paragraph or
conversationally, it changes the
sense it can give. To illustrate this point: two users can say 'pets', where the
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


29

digging themselves out of the hole we had dropped them into! Their assessment
of the
situation and also each other's
presented

selves assisted them in this, since it was by no
means inevitable at the outset of the situation that they would not get on famously with
one another. They had to discover during the interaction whether they

really did have
grounds for getting better acquainted (or not).


Using shared interests to matchmake residents in a neighbourhood and then via the
activities of personal (software) service agents introducing those agents puts the Living
Memory System in

the positions of 'formally' making an introduction as [Goffman, 1963
#351] (page 120) notes,
' An introduction, even more than acquaintanceship that
develops informally, ought, it is felt, to have a permanent effect, placing the introduced
persons forever

after in a special and accessible position in regard to each other.
'


Living Memory agents as a third party are therefore placing local residents
in a special
relationship

to one another. One which we can imagine that there are perfectly good
reasons for
wanting to avoid as well as to accept. Moreover the larger point to be made is
that if a Living Memory system created a proliferation of these kinds of introductions
then it might perhaps ironically result in its users going to great lengths to evade eithe
r
the system or one another. And since they, by definition, live on
proximate

territory, they
would have great difficulty
avoiding

one another since they are likely to see one another
at the shops, walking the dog, out for coffee, at church etc., then the
sensible choice
would be to avoid the matchmaking agent in the Living Memory system.




Agents amending the user profile on the basis of their inferences


As we noted in the previous section on the Acquaintance Model element of the user
profile, the Livin
g Memory system's software agents are not meant to be simply the
passive readers and matchers of a user profile. The software agents pro
-
actively manage
the user profile (see Box 3 & Box 4).


Box 3
Extract from LiMe Deliverable 4.1 Agent making inference
s on the basis of
disinterest of user while using the Memory Mapper


The pro
-
active element (agents making sense of what users do
-

see particularly box 4)
has been continually emphasised throughout the LiMe project. In this sense the Living
Memory system
would do some of the work of LiMe users by for instance monitoring





first has a cat and the second caged
-
birds, and the second hates cats and first hates to see birds caged (even
though its cat sometimes kills birds).

In future versions of the system we plan to record such actions (removing an item of LiMe
con
tent from view) and use the deleted items as models for what the user is not interested in.
This will help the agent to understand more clearly what the user is looking for and in which
context the specified keywords are of interest to the user.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


30

whether users opened material on the LiMe system which matched their user
-
profile list
of interests. If a user appeared (to the software agent) to
never

be opening items of
content which ma
tched a particular interest then the agent would ask the user whether
they wished either to remove the interest from their list of interests or whether they
needed help in finding items in the LiMe system which corresponded to the interest. Or
equally as n
oted in Box 3, if a user continually discards items which a software agent
considers to be of interest to the user on the basis of the user profile then the agent would
need to 'wise up' and amend the user profile.
14




Box 4
-

Extract from LiMe Deliverabl
e 4.4, the
explicit

user profile and the
deduced

user
-
profile


In terms of what was implemented in the Living Memory prototypes we were unable to
trial or evaluate deductive changes made to user profiles by software agents on the basis
of the log of actio
ns with the system. Nevertheless we have spent time reflecting in the
light of our action research and earlier classic CSCW research by Lucy Suchman
[Suchman, 1987 #226] on photocopiers on what an agent can know of the social situation
of its user. And of
course we have turned to the ultimate expert on deduction, Conan
Doyle's fictional detective 'Sherlock Holmes'. To begin with Holmes' technique:


--




14

In other words the user

is getting the equivalent of 'junk mail' from the software agent(s) and is binning
most of it so the software agent which has access to the user's binning activities and the contents of their
'trash' can amend what it sends. Though arguably as a semi
-
aut
onomous agent whch is part of a
community of such agents it may also have obligations to dissemination agents to receive their mail even
though most of it is thrown away by its user.

A user of a connected community is a community member allocated a
user space

within the
system. The user space holds
memory content

owned by the user and a
user model
. The
memory content holds private memories represented by means of hypermedia documents
suitably structured according to the taste of the user. The model of the user
is structured in
terms of:



User History



a set of events holding a log of the actions made by the user during
interaction with the system;



User Profile



a set of information including personal details, preferences, interests and
needs of the user, each s
pecified by set of suitable concepts, criteria and constraints;



Acquaintance Model



a set of relations describing the social links of the user with other
members, places and activities in the connected community.

Information in the user profile and the ac
quaintance model are either described and updated
explicitly by the user through interaction with the system, or deduced implicitly by interpreting
the actions of the user recorded in the user history. In this way, we combine general
descriptions of the us
er
-

according to the knowledge held in the profile and acquaintance
model, with specific information about the user
-

obtained from what the user actually does.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


31


From Chapter 1, The Sign of Four by Arthur Conan Doyle.


"Not at all," I answered earnestly. "It is of th
e greatest interest tome, especially since I have had
the opportunity of observing your practical application of it. But you spoke just now of
observation and deduction. Surely the one to some extent implies the other." "Why, hardly," he
answered, leaning
back luxuriously in his armchair and sending up thick blue wreaths from his
pipe. "For example, observation shows me that you have been to the Wigmore Street Post
-
Office
this morning, but deduction lets me know that when there you dispatched a telegram."


"Right!" said I. "Right on both points! But I confess that I don't see how you arrived at it. It was a
sudden impulse upon my part, and I have mentioned it to no one."


"It is simplicity itself," he remarked, chuckling at my surprise
--

"so absurdly simp
le that an
explanation is superfluous; and yet it may serve to define the limits of observation and of
deduction. Observation tells me that you have a little reddish mould adhering to your instep. Just
opposite the Wigmore Street Office they have taken up
the pavement and thrown up some earth,
which lies in such a way that it is difficult to avoid treading in it in entering. The earth is of this
peculiar reddish tint which is found, as far as I know, nowhere else in the neighbourhood. So
much is observation
. The rest is deduction."


"How, then, did you deduce the telegram?"


"Why, of course I knew that you had not written a letter, since I sat opposite to you all morning. I
see also in your open desk there that you have a sheet of stamps and a thick bundle

of postcards.
What could you go into the post
-
office for, then, but to send a wire? Eliminate all other factors,
and the one which remains must be the truth."


"In this case it certainly is so," I replied after a little thought. "The thing, however, is,
as you say,
of the simplest. Would you think me impertinent if I were to put your theories to a more severe
test?"



What relevance might Sherlock Holmes have for software agents? Sherlock Holmes, as
he explains his method, uses both observation
and

deduct
ion. He is able to
see

the
reddish mould on Watson's instep. At some point earlier when Holmes has been out in
the streets of his neighbourhood he has seen not only that the pavement has been lifted
but that pedestrians are having to step in the red earth
to pass by. Added to that, apart
from knowing Watson's habits well, he has been watching what Watson was doing all
morning from his position across the room. Holmes uses his common knowledge of what
people go to post offices for and then knowing that Watso
n has not been writing a letter
he deduces he must have gone to the post office to send a telegram.


What is clear from Holmes' method is that there is far more active
-
observation required
than there is deduction done. He has to keep a watchful eye on his

neighbourhood, and
his companion Watson, and to make his deductions he has to remember the relevant
events (since there doubtless a multitude of irrelevant events during the morning). If we
substitute for just one moment (with apologies to Conan Doyle and

Holmes) a software
agent detective (sniffer even!) and imagine that Watson were using a networked
computer rather than writing on paper. The software agent detective might have
remarkable powers of deduction and yet without the first observation: reddish
mould on
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


32

Watson's instep, it would not be able to make the subsequent deductions. And it would
also lack the ability to see the scene outside the post office to which it could attach the
significance of the dirt on Watson's shoes. And ultimately the point
that Holme's makes is
that he is also able to see what Watson did
not

do, which was to use any of his postcards
or stamps.


Lucy Suchman's [Suchman, 1987 #226] work on photocopiers brought out the manner in
which when photocopiers attempted to diagnose a f
ault they often caused all kinds of
problems for the user because their situated sense of what was happening was
asymmetrical. In other wiords the photocopier had only a very weak sense of the situation
compared to the human technicians. Her parallel point

was that the users of the
photocopier struggled to understand the photocopier's 'reasoning' because compared to
commom human members the photocopier and the humans lacked a reciprocal
perspective on what was going on.


Successful work with inference
-
makin
g software agents has been reliant on their
'watching' what the users does via actions on a keyboard producing ASCII text in
response to the organised features of computer programmes (i.e. [Rhodes, 2000 #397]).

Other examples of the "observations" through

which software agents can deduce interests
are through inspecting the kinds of events generated by an internet browser (bookmarks,
history files) [Moukas, 1997 #399] and by consulting other agents. Yet one of the key
element of the Living Memory system is

that it would not only be based around keyboard
and screen PC interfaces, indeed as manifest in LP4 it has worked toward merging
system access points with the existing furnishings of public places (i.e. café tables,
noticeboards and bus stop display areas
). Simple input/output mechanisms such as tokens
(see Fig .1) might be used to allow people to become carriers of information and to assist
in both simplifying the movement of system content and the sharing of that content (see
'Living Interfaces' and LP4
final report). In this way the system itself might arguably
acquire a distributed sense of what is happening in the neighbourhood rather than only at
the PC interface.


However the task of "watching" the user becomes massively complex as the user shifts
fr
om role to role (parent to café customer), place to place (kitchen at home to aerobic
class) and channel to channel (telephone to postcard). For a personal software agent (or at
some 'higher' level the LiMe system) to have
sufficient

grasp of interactions
that happen
on paper, face
-
to
-
face, by telephone or in any other way off the LiMe electronic network,
then it has to have ways of observing these events and checking that its observations
match its users versions of those events (because even Sherlock Holm
es made mistakes
in his investigations). To user a very simple example which QMUC have alerted their
partners to from the very beginning, if a token is given by one user to another this
transaction is not observable to the system (unlike the transfer of an

e
-
mail between one
account and another). If a bouquet of flowers is given by one person to another then this
is not detected nor should the person throw the flowers in to the gutter in disgust or anger.
Yet many of these 'non
-
system' events are highly rel
evant to amendment and updating of
a Living Memory user profile by a software agent.


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


33


To summarise then, user
-
profiling involving lists, matchmaking by software agents and
implicit observations of users by software agents though undoubtedly innovative rai
ses 2
significant issues:


1

Bearing in mind that Living Memory is routed and rooted in a territory, (it serves a
bounded geographical neighbourhood), the matchmaking of residents on the basis of
their user profiles and their subsequent introduction by their

agents/LiMe system may
or may not result in people avoiding using LiMe so that they do not have to avoid
unwanted acquaintances in the street. In other words LiMe could cause an
acquaintance overload on the basis of shared interests.

2

Implicit adjustments
of user profiles by software agents remain problematic given
software agents limited observational capabilities of members' social situations.
15





Profiles of user's everyday lives for Philips Design (WP1 & WP5) Community
Narratives


As is no doubt becomi
ng clear to the reader of this report, user profiling existed in several
different formats in the Living Memory Project. There were place
-
based profiles of users
for Domus Academy’s concept design, there were networked diagrams of acquaintances
for Imperia
l College’s Acquaintance Model (AM) and lists of a person’s interests for
Imperial College’s matchmaking by software agents. These latter two forms of user
profiling were co
-
utilised by software agents since they were to make implicit
judgements on the bas
is of who
-
knows
-
who in a neighbourhood and what they do
together and seperately. In this final section on user profiling by QMUC for Philips we
see how they translated Imperial College’s forms of user profiling into their ‘slice of
LiMe’.
16


Fig. 14.
A Livi
ng Memory Event: 2 women get to meet, help and know one another
through losing and finding a cat.





15

By coincidence the Microsoft Word helper has incorrectly (to me, thoug
h not to the software)
renumbered this list 3 times in response to my actions on the keyboard within its programme structure, it is
presumably seeing this numbered list as a continuation of the list that concluded the previous section. A
classic example of

an ordinary human user struggling with a deduction made by a computational artefact.

16

‘Slice of LiMe’ was a
compromise

between the visions for a ubiquitous, many
-
headed and
comprehensive Living Memory system which would realise the ambitions of its desig
ners and an
assessment of which elements could be coded, built, assembled and tested within the project’s 3 year life
cycle.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


34




An ongoing ‘just so’ situation for Living Memory has been that of the Living Memory
system being able to find a lost cat.
17

It finds the lost cat not only

through the system but
also by bringing together the ‘real knowledge and memory’ of ‘community members’
('Living Interfaces', page 6). In the story illustrated in Fig. 14 Mrs Irvine adds a missing
cat notice to the Living Memory system. Meanwhile Liz also

adds an advert for a
barbecue that she is organising. Later at a bus stop with a Living Memory public viewing
point (i.e. electronic display screen) she sees Mrs Irvine’s advert about her lost cat. After
her barbecue Liz spots the missing cat but is appar
ently unable to apprehend the feisty
feline which makes off. Since there is a LiMe interface in her local café Liz is able to
recheck the missing cat notice and get in touch with Mrs Irvine. Presumably they return
to the place where the feisty feline was l
ast seen and somewhat surprisingly for a cat


it
is still there. Having returned home happily with her captured kitty Mrs Irvine sends a
thankyou note (either from, or it is received at a local café). The story has a happy ending
when Liz, Mrs Irvine and
the cat settle down to have dinner together. The Living
Memory system in turn through having recorded all these individual system events
threads them together (see fig 17
-
19).


How the story of the lost cat relates to user profiling becomes apparent if we
move on to
the specification of user agents (fig. 15) and place agents (fig. 16). Liz and Mrs Irvine as
users of the Living Memory system have agents at their personal service and other agents
who they or their agents interact with at certain places in the

story (the bus stop, the café
and possibly the barbecue and at dinner). When Mrs Irvine places her 'lost cat' ad then her
user agent adds the action to her profile as part of her history and also logs the place and
time it was added (in a similar fashion
to the 'read' event as it forms a LiMe node in
Fig.20).



Fig. 15.
A Living Memory Event: specification of LiMe user agent





17

By 'just so' we mean, a story that is convenient for a designer's conjecture about what
might

be but whose
details are not (yet)

subject to empirical constraints (see [Lee, 1993 #427] pp27
-
28).

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


35




Mrs Irvine's 'user agent' negotiates with place agents (see fig. 16) to reach agreement on
the places in which her lost cat ad w
ill be displayed. In the Philips scenario this is the bus
stop. Though it may be that the ad is displayed at the bus stop in response to the detection
of Liz (via a personal token she is carrying) whose user agent indicates that she likes cats.
However sho
uld it be the first option, and Liz sees the missing cat notice without having
entered a 'read' command into the LiMe system, then although the missing cat notice has
been read by Liz the system is
unaware

of this event. Liz's user agent does not record th
e
activity and the system does not record the activity. It is to all intents and purposes an
invisible event for the system.



Fig. 16.
A Living Memory Event: specification of LiMe place agent





In terms of actually finding the missing cat the Living M
emory System relies, in the end,
on another human finding a cat and seeing that it might be 'a lost cat' or indeed 'the lost
cat' (if they have seen an image of the missing cat already). How a human sees a cat as
'lost' as matter of finding a cat to be ou
t of place is not something QMUC have
investigated. Seeing an image on a poster or displayed on an eletronic noticeboard is
certainly a potential part of it (though if Liz happens to 'read' a paper poster taped to a
lamp
-
post then this reading event is, on
ce again, invisible to the system). Should Liz
when she spots the cat not bother reporting her spotting via the system to Mrs Irvine, then
the system never learns that the cat has been seen. Equally when the cat is eventually
found by the two women this fi
nding has to be reported to the system since it does have
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


36

'senses' to register this event either. There are other ways of making sure the system
'knows' about these events and can add them to a user's history and amend their profile,
and acquaintances acco
rdingly, such as the system querying the original cat
-
loser 'have
you found your cat yet?' Although this may be particularly heartless if the cat has been
found squashed flat on the road or/and its owner has tried to "forget" about their loss. A
human can
make judgements about the sensitivity of asking
-

can an agent? However the
point is also that the system to make sense of losing and finding cats has less resources
available to it than its owner and will accomplish its task as part of the combined
search
ing, reading, finding and reporting activities of the loser, cat and finder.


If we now move on to the description of Living Memory's 'memory' bearing in mind that
many 'events' which are relevant to the story of the lost cat are missing from the system's

history then the potential of the Living Memory system being able to 'thread' together
stories, narratives, local news or even perhaps 'memories' has to be treated with a high
degree of scepticism.


Fig . 17
Extract from Philips Design Description of 'M
emory' in the Living Memory
System
-

relations between nodes.

The network structure as described in Fig. 17 and then expanded upon in fig.s 18,19 & 20
is necessarily incomplete when it is treated as an 'internal' network which is held in the
'memory' of th
e system. If we allow all the actors
-
in
-
action with their varying projects
back into the network (and this is still an inevitable vastly reduced list!):


Liz
-
at
-
a
-
bus
-
stop,

Mrs Irvine
-
looking
-
for
-
her
-
cat,

a cat getting hungry and begging,

photocopied p
ictures of her missing cat secured to lamp
-
posts,

the street pattern that brings Liz to see the cat,

Liz's ability to match the image of the 'missing cat' to this begging cat,

a telephone number,

turn
-
taking aspects of telephone talk,

touchscreens,

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


37

to
kens,

cafes,

café tables,

and editors of local newsletters in search of a story.


Then the story as it is told at the beginning is narrated in a somewhat different form by
either Mrs Irvine or Liz or perhaps most likely by the local newspaper. Yet even th
ose
tellers of the missing moggy tale do so by bringing to bear multiple resources,
knowledges, sense of relevancy on their brief narrative [Sacks, 1987 #368]. If we turn to
the process whereby the Living Memory system both stores a threaded web of 'nodes'

and
then it produces a story from its web by tracing a connection (see fig. 19), then we can
see support for our ethnographic worry that a Living Memory system's history of its user
may not be sufficient to grant it rights to alter user profiles or have t
he 'entitlement' to tell
a story
18
. In part this is because even the most powerful computational artefacts have, as
Suchman (1987, p170) neatly puts it, 'severe constraints on their access to the evidential
resources on which human communication routinely r
elies.'


Fig. 18
Extract from Philips Design Description of 'Memory' in the Living Memory
System
-

A 'crystallised' network of system 'memory'.



Contrary to what we might expect from certain notions of panoptic technologies (i.e.
CCTV, networked registra
tion databases for smart cards etc.) even distributed agent
based connected
-
community systems sense only tiny fragments of what human members
sense (as skillful learned hearing, kinesic, seeing, tasting, touching organisms). From its
incomplete fragments t
he Living Memory system would then have to work like
archeologists with a few shards of a pre
-
historic earthenware pot to tell a story about a
long lost feasting event. Except the additional problem is that unlike an ordinary member



18

[Sacks, 1992 #176] makes an important distinction to do with entitlement to tell stories about events, so
that should Mrs Irvine or Liz decide to tell about their finding the missing felin
e; the ownership,
plausibility and coherence of their story turn on their having had firsthand experiences of the events in
question. As [Lynch, 1996 #137] page 280, adds, 'This relation between experience and storyability runs
deeper than the question of

how stories draw on a fund of experiences; it also has to do with an
attunement
in the course of an experience

to the possibility of later telling a story about the experience. Witnessing and
storyability are, in this sense, methodologically interwoven.'

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


38

the Living Memory syste
m does not look at unfolding events to see what might be
retained for the production of a story at a later point (see previous footnote).




Fig. 19
Extract from Philips Design Description of 'Memory' in the Living Memory
System. Illustration of the threa
ds being 'pulled out' of the weave to produce fig. 14



As [Rhodes, 2000 #397] notes in his work on software agents which create
personalized

newsfeeds for users 'the "sensors" are all looking at the newswire not the user' (page 2).
For Rhode's remembranc
e agents the personalisation comes not from a static user profile
but from watching the 'user and the user's current context'. However the key point to note
is that the agent is able to watch the user because the user's work
-
setting is at a keyboard
using
web
-
browser software. In other words the user is highly configured and constrained
by having already engaged in a kind of 'obligatory point of passage' [Latour, 1993 #108]
for writing with a word processor by
passing through the PC set up
. The agent can
re
gister sufficient changes in what the user is doing because the setting is tightly
organised for doing computationally relevant tasks. For Living Memory as we noted at
the outset we have unusual users who ideally are not to be found typing articles,
spread
sheets or business plans or presentations in an office at a keyboard, screen and PC
interface. For Living Memory we have users who are queuing at bus stops, sitting at
tables in cafes and sometimes looking for lost cats.



Fig. 20
Extract from Philips Desi
gn Description of 'Memory' in the Living Memory
System. The inside of a single node in the Living Memory web.


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


39



A final point we would like to make in relation to user profiling is that if user and place
profiles accrete complex and lengthy histories as
time passes, they will have to 'forget'
various events or else user profiles may end considerably longer and more complex than
those which QMUC produced from interview data (see Appendix 1). Yet the user profiles
should be inspectable and manageable by loc
al people without expertise in the use of
computers or a great deal of expertise in existing software systems.


To summarise then, user
-
profiling involving imaginary scenarios, threaded events, and
evolving user profiles though undoubtedly innovative rais
es significant issues:


1

User profiles need to be trialed in actuality rather than through scenarios, since even
with our action research and media archetypes approach we were unable to develop
an appropriate hybrid to investigate the evolution of user pro
files where LiMe system
events, actions & histories were involved.

2

If user profiles are to be disengaged from the typical PC software interface (where a
user may have different profiles according to what product they are using
-

thus being
several users o
n one keyboard), then these profiles are either going to have to be
simple but multiple or in some other way heterogenous to cope with the different
roles that a member in a neighbourhood takes according to which community of
practice they may be a member
of (and there are likely to be a number of different
communities
-

i.e. as part of pet owners, volunteers, parents at a particular school, a
café customer, of which one embodied person can perform roles in).

3

The repurcussions of the impaired senses of comp
utational artefacts have to be
realised in terms of their ability to remember (as action) rather than its ability to store
(as accumulation). It is unlikely that they can tell a simple story about how they saw
two women finding a cat unless we change what
telling a story is (see [Button, 1994
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


40

#308] for an equivalent and better stated argument on computers having a
conversation).

4

The Living Memory system has to be able know what it can and cannot forget
according to its situation as an accountable artefact (
see [Lynch, 1996 #137] on the
job of remembering.)



LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


41

Concluding Remarks.


Our final report has raised many questions about the implications of new kinds of user
profiling. Many of these are to do with Living Memory treating its users in a highly
irregular

manner compared to the typical configuration of human actors for
computational artefacts [Woolgar, 1993 #304]. Since LiMe users are allowed be more
polyvalent and occupy multiple networks [Law, 1994 #126]. The question then is how
can a Living Memory syst
em know enough about local residents to be 'personable' to
them.


To re
-
iterate the questions we have raised along the way:


1. If the Living Memory system uses acquaintance models then how are mutual
agreements reached on what the nature of an acquaintanc
eship is between 2 residents? An
acquaintanceship which will vary over time and according to situation and according to
which member is describing it.


2. If there is an interconnected web of acquaintance models created by the Living
Memory system for all
residents in a neighbourhood, who is able, willing and/or
entrusted to inspect, utilise, regulate and validate the links?


3. Relatedly, how are the
inferences

about who knows who and in what capacity made by
the Living Memory system to be rendered account
able. Any such inference is normally
defeasible by the persons referred to. In other words, users should have the primary
entitlement to say who they know and in what capacity.


4. The matchmaking of residents on the basis of their lists of one
-
word intere
sts in their
user profiles and their subsequent introduction by their agents/LiMe system may or may
not result in people avoiding using LiMe so that they do not have to avoid unwanted
acquaintances in the street. Bearing in mind that Living Memory serves a

bounded
geographical neighbourhood and that each person does not want to every other person
that shares an interest with them.


5. Adjustments of user profiles by the deductive powers of software agents remain
problematic given software agents limited
obs
ervational

capabilities of members' social
situations. They have the capabilities, metaphorically, of Sherlock Holmes solving cases
while stored in a locked and windowless box with an ASCII ticker tape being fed in
through the letterbox.


6.
User profiles
need to be trialed in actuality rather than through scenarios, since even
with our action research and media archetypes approach we were unable to develop an
appropriate hybrid to investigate the evolution of user profiles where LiMe system
events, action
s & histories were involved.


LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


42

7. If user profiles are to be disengaged from the typical PC software interface then these
profiles are either going to have to cope in some way with the different roles that a
member in a neighbourhood takes according to whic
h community of practice they may be
a member of at any particular time. In other words an actual human resident of our study
area going about their daily activities is a member of multiple communities of practice
not one 'local community'.


8. The repurcus
sions of the impaired senses of computational artefacts have to be realised
in terms of their ability to remember (as action) rather than its ability to store (as
accumulation) and relatedly the situated production of stories has to be taken into account
i
f the Living Memory system is going to produce stories that are in any way meaningful
to local people.

9. The Living Memory system has to be able know what it can and
cannot

forget
according to its situation as an accountable artefact (and this last one so
unds like a zen
master's puzzle!)
LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


43

Appendix 1. Anonymised User Profile Produced by WP2 Year 1, (excerpted from
Deliverable 2.1)


1. PERSONAL PROFILE


Name:




Laura Kyle

Gender:



F

Age:




34

Ethnic background:


Scottish, white

Education:



6 ‘O’ levels,
4 ‘A’ levels, Diploma in Business Studies


2. HOME ENVIRONMENT:


Family situation:



Lives with husband and one daughter (aged 10)


Role towards family members/

Primary caregiver of child, shares responsibility

Family Responsibilities:

of running househ
old with her husband, part
-



time work contributes to household income. Sees sister


and her children every week (share childminding



sometimes), visits her mother (once a week) who lives


locally.

Housing:




Owner occupied house

Where:




Clerwood


Type:


Typical ‘second’ house for houseowners, well
equipped with conservatory, small study, garage
and small garden

Size:

3 bedroom

Restraints (in relation to role/time):

As well as domestic responsibilities, spends time
with her daughter doing things
(takes her to her
clubs/ sports activities), does voluntary work at
school (helping children with reading in the library),
on church social committee.


3. ATTITUDE TO TECHNOLOGY:


Technology awareness:


Limited
-

mostly directed by children (who

know more

about how to use the systems they have)

Computer literacy:


Limited
-

has good typing speed. Uses



computer much like a typewriter

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


44

Why (use computer?):


Sometimes supervise children, Occasionally write
her own personal and business letters, CV etc

Wh
at (systems used?):


Microsoft Word, Excel (rarely), encyclopaedia
software for children’s education (mostly the
children use and she watches)

Where (computer is used?):


Study at home

When:


When children go on it mostly
-

two nights a week,
occasionally

on weekend (to write letters)

With whom:


With children


4. MEDIA/COMMUNICATION/INFORMATION SHARING

(A) IN RELATION TO SITE


Which? (Mode of Communication

Face to face, phone, letter (school sends to her),
/Type of Information sharing):

newsletter


Wh
o with:


F
-
to
-
f: with other parents
-

a lot and after school
club (ASC) leader occasionally teacher and
headteacher/ phone other parents, get letters from
school secretary

Where
:

F
-
to
-
f with other parents at the school gate/ F
-
to
-
f
with teacher/headtea
cher in corridor or classroom

What purpose:


School gate: to chat, gossip, to see how a parent is,
how child is (behaviour, health, education
problems, stories about children), talk to teacher
about any concerns parent has for child’s education,
talk to A
SC leader about days I’ll be using the after
school club times to pick up etc & things I need to
give child to bring.

Frequency/intensity:


Other parents: 4 times a week before and after
school, sometimes twice a day

Attitude about communication

F
-
to
-
f a
t school gate is where all informal
/information sharing:


information about school is exchanged
-

stories


about children and their parents. Get all the I


information needed about school and events etc


through the newsletter.

How could communication/


Between other parents
-

more parents could be
sharing information be improved?


involved with Parents association meetings and

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


45


fundraising. Also would like to hear other parent’s
experiences of their children & education, if they
have any problems. Wo
uld like to talk to teachers
more (but difficult to have teacher to find time for
this).


(B) MEDIA/COMMUNICATION/INFORMATION SHARING IN RELATION TO
LOCAL AREA:

Mode of Communication/info

F
-
to
-
f/phone/read posters

Who with:


Librarian, ASC leader, Sp
orts Clubs, Leisure
Centres

Where:


Library, Sports Club noticeboards

What purpose


Find out and share info about planning leisure
activities for child, ask questions about how recent
posters are on local community noticeboard, read
the Corstorphine news

Frequency/intensity:


Every 1
-
2 weeks. Spend half hr to an hr (time
includes making enquiries by phone, visiting
library)

Attitude communication/

Library is good source of local info. Not
information sharing:


always easy to visit during the week. Ca
n take


awhile to sort through all info to find



something suitable for my child’s age range.

How could be improved?


Would be easier if could get all the
sports/leisure/local events info in the one place and
if that info was organised into age groups.

Save a
lot of running around and sifting through irrelevant
info.


5.
MEMORY:


(A) COLLECTING IN RELATION TO SITE

What’s (important to keep):


Newsletters, school reports, some artwork

Why:


Newsletters
-

To know about current/topical events
taking plac
e at the school
-

keep for a month then
throw out. Artwork: Try and keep pieces either
parent or child likes to remember when child gets
older, or to display. School Reports
-

record of
child’s educational development and achievements.
Keep for parent
and child for the future.

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


46

Where (keep things you collect?)


Newsletters
-

on pinboard, school reports
-

in
document folder, Artwork
-

on wall, pinboard or in
attic.

When:

Newsletter
-

every week (or until out of date or
when parent has a ‘clean out’). S
chool reports
-

get
twice a year. Keep all of them. Artwork
-

keep
items say once every 2
-
3 months.


(B) COLLECTING
-

IN RELATION TO LOCAL AREA

What do you keep (eg diary):

Photos, diary (personal thoughts), scrapbook (things
which connect with me
-

loc
al events or issues, a
good photography, articles about things I’m
interested in and photos/items on people I know
-

from local newspapers, magazines and
national/international magazines and papers)

Why:

Photos
-

to remember moments and special
occasions

for the future
-

so I can remember what
my child looked like at 2 yrs and 5 yrs old etc (and
how parents and other family have changed.), diary
-

Sometimes to write in my personal thoughts in
special (good and bad) things that have happened
that day/week/
month

Where (keep things you collect?)


Photos
-

in albums in study and loft, Diary
-

in
exercise book (hidden in bottom drawer),
Scrapbook
-

in study.

When:

Photos
-

after holidays, Christmas, weddings, diary
-

every month of so record things, scrapbook

-

every
1
-
2 months find something to put in

With whom:

Share photos and scrapbooks with family


6. OFF SITE:


(A) WORK

Type of work:

Senior Loans Officer at Bank of Scotland

Colleagues relationships
:

Close friends with 2 other women at work

Spare ti
me Hobbies:

Needlework, reading (not much time for own


hobbies, spends time with daughter at her


activities/sports).

Sports:


Badminton

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


47

With whom


Plays badminton with other parents (some on
Parents’ Association) at local school

Intensity

Plays ev
ery Thursday night


(B) SOCIAL CONTACTS:


Clubs:


N/Ap

Activities:


On fundraising committee with Parents
Association). On Church Social Committee, does
voluntary work at local school (reading for children
at school library).

Relationship with club mem
bers?


Friends with the other women at Parents
Association (about 14 in all, but close friends with
4
-
5 of them).

Effort to become member?


No effort to become a part of Parent’s Association,
but some people think it’s ‘cliquey’ (a closed group)

Intensit
y of contact

Friendly chat with the other women at Parent’s
Association and Church Social Committee.
Sometimes sit around and ‘philosophise’ about
education and religion, but mostly talk about
everyday things (the children, work, the household,
husbands,
socialising, cooking, school, education).
The 4
-
5 parents very friendly with, go to each
other’s house every month of so for lunch or cup of
tea.

Friends/Mutual interests activities:
School, children, badminton, school events.



7. ON SITE:


(A) ACT
IVITIES

Why visit the site/What about?


To drop or pick children up from school or ASC
(depending on hours of part time work). Join other
parents at coffee mornings (Parent’s Association),
attend parents Association meetings (every month)

When?


Most mor
nings and afternoons/ASC
-

early
evenings

Frequency?

Four times a week (on the 5
th

day sister organises
picking up children)

LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


48

Together with whom?


Usually visits on own. Occasionally walks with
other parents between home & school, if happen to
meet on the

street/at school gate.

Effort keep contact with site visitors?

Little effort when meeting at school gate.
Sometimes difficult to meet for coffee mornings,
depends on how busy parent is.

Intensity of contact


Usually just chatting, but with some of th
e women
parent is closer to, contact is more intense.


(B) HISTORY

Number of years/months/days coming:

Since child started school 4
-
5 years ago.


Lived in Corstorphine for 25 years

Change over time in activities?


More involved with child’s education wh
en she first
started school (helped in the classroom).













LiMe D2.5


Case Study of User Profiling

LiMe Doc. No/QMUC/EL+KB+AW/200005


49




REFERENCES


Agostini, A., Michelis, G.D. and Susani, M. From User Participation to User Seducation
in the Design of Innovative User
-
Centred Systems Digital Creativity Journal XXXXX


Bl
omberg, J., Suchman, L. and Trigg, R.H. Reflections on a Work
-
Oriented Design
Project Human
-
Computer Interaction 11, 1996, 237
-
265


Carroll, J.M and Rosson, M.B. Getting Around the Task
-
Artifact Cycle: How to Make
Claims and Design by Scenario ACM Transact
ions on Information Systems 10(2), 1992,
180
-
212


Hughes, J., King, V., Todden, T. and Andersen, H. Moving out from the control room:
ethnography in system design, in Proceedings of CSCW ’94 429
-
439


Moores, S. 1990 `Texts, Readers and Contexts of Reading:

Developments in the Study of
Media Audiences', Media, Culture & Society, 12(1), pp. 9
-
29.


Rogers, Y. Reconfiguring the Social Scientist: Shifting from Telling Designers What to
Do to Getting More Involved, in Bowker, G.C., Star, S.L., Turner, W. and Gass
er, L.
(eds.) Social Science, Technical Systems, and Cooperative Work. Lawrence Erlbaum
Associates, London, 1997


Shapiro, D. The Limits of Ethnography: Combining Social Sciences for CSCW, in
Proceedings of CSCW ’94 (Chapel Hill, NC, Oct ’94) ACM Press, 41
7
-
428