Greg Detre
10:30 on Wednesday, October 02, 2002
Present:
Greg, Napier, Ted, Andrea, Win
Andrea
chose those to show the disjointedness and lack of consensus in the field
what role
does context play in everyday?
extension to technology?
what does comp. really do for us?
can we interact invisibly and have control?
can we feel served and safe?
agenda:
define context
representation of context
�Context: any information that can be used to characterize the situation
of entities (i.e. whether a person, place or object) that are considered
relevant to the interaction between a user and an application, including the
user and the application themselves. Context is typically the location,
identity and state of people, groups and computational and physical objects.�
four categories of components:
1.
interpreters
2.
aggregators
3.
services
4.
discoverers
context-aware brings together ubiquitous computing and tangible
interfaces
embodiment � participatory in the world
Heidegger � the meaning is already (implicit???) in the world � you just
have to find it
�for Heidegger, the primary question is not �how do we assign meaning to
our perceptions of the world?� but rather �how does meaning of the world reveal
itself to us through our actions within it?��
context emerges as a solution for the business of everyday �
ethnomethodology
�approach to social analysis which explains the orderliness of social
conduct not in terms of abstract theories, but rather as the practical
achievement of members continually working to render the world sensible +
interpretable in the course of their everyday actions�
it should
be partly about allowing more of the interaction to recede into the background
defines context as everything that affects the communication
channel/interaction between the user and the computer
context doesn�t benefit from Moore�s Law
context is shared � whether user/application or user/user
blackboard
sensors � announce and listen
missing the
idea that there are expected situations, models of task and circumstance that
is crucial
continually in the (unexpected/uninformed) world of sensor-effector
widgets
completely separates representation from the application � that completely
throws away the context
Ted: where�s their representation of �meeting� or �dinner�?
don�t they lump together gathering + representation together � perhaps
they see the widget as doing more, as kind of the first half of a two-part
blackboard
i.e./e.g. perhaps you would have a �meeting widget�?
have a hierarchy of widgets
so much of the conversation at conferences with parallel sessions is
about where things are
I like Dey�s conference example � it�s kind of like college � I would
definitely want an object-orientated approach to navigating classes (the main
objects), synthesising everything together
cheat sheet of important facts � make explicit the structure that
organises things (e.g. all the odd-numbered buildings are on this side of
Killian Court) � giving you the information when you need it
I would
never give up control by taking notes on someone else�s system and just trust
that they�ll be available online when I get home
different levels of trust � accountability, but also grows with
experience
I only trust my Palm because it synchronises with my laptop � and I can
provide the accountability by back-ups
saving web pages to my hard disk is an encumbrance
it allows
you to compare different types of information to each other
is Winograd
focusing too much on implementation???
no � perhaps you have to bear in mind that the article only draws attention to half the story, because the interesting aspect for context-aware people of the blackboard approach is how you gather the elements of the blackboard together???
then you have to posit another level of relations (like a gatherer-widge)�
how would
you model (say) the Action History in a desktop application?
the blackboard makes more sense for this, because you could have a
cross-application Action History
you wouldn�t have to design it ahead of time � you could add it
retrospectively, if all the actions are already posting to a blackboard
when would
you use a networked model?
it�s good because it allows thin clients, e.g. your PDA doesn�t need any
sensors if it knows where to query for that information
announce-and-listen
isn�t failsoft
the MIT
group focus on very targeted situations
you can�t necessarily tell when you�ve stepped outside (swtiched
context) that targetted situation unless you have a model that encompasses it
context-switching is just unbearably difficult
very few of the MIT demos are multi-person
do you want
information in your mind or externalised???
you need an
architecture/backbone that decides what subset of the information to present,
and learns from my feedback to shape that subset � the blackboard model is one
such architecture
we have to get away from the designer deciding what information to put
in the system
it�s more flexible, if you want to present information organised in a
new way that hadn�t been anticipated
it allows you to impose (acetate-like) layers of priority over the huge
dataset
this way, you don�t have to trust the designer
how do you add to/modify your system without breaking it?
where do I put the knowledge? where do I put the decisions? how do I
change that?
discrepancy
between what people like to do and what they ought to do
should the profile that the system builds of you simply react to your
preferences, or should it (to some degree) try and influence your preferences
at the moment, the system is not smart enough to be trusted with any
sort of priority over your decision
but how do they react � to your explicit goals, or to your behaviour?
context-aware systems don�t ever have to make the big decisions � they
do shape them in terms of (say) the way they present that information
but let�s say you�ve defined your user profile in some way, but you�re
acting differently � it might want to signal that discrepancy
should we
try + compare the diagrams each author presents?
the system
is never going to be right � how do you design with that in mind? how do you
cope with system failure?
do we want
to do more papers from the same journal issue?
y
let�s spend
some time actually creating a diagram for a given system (for an example)
e.g. a life-work scenario
try and
refer to these papers
make
suggestions about what to read next week
napier will
think about GIS
wynne will
try and think about the discrepancy between what you want and say/ought to want
e.g. life-work balance
context outside CS?
ted will
try and think about how his lab differs from other people�s CA stories
in a world
where we can�t trust anything, what can we do?
I want to
be positive, and think �when can this stuff work?�
how close
is the analogy with a GUI (signals and slots, widgets etc.)???
do you ever
want widgets to be able to react, and vice versa???
what about
task???
is context-aware computing pretty much just engineering?
winograd � event heap/context memory = short vs long-term memory???
ted: pervasive (back-end) vs ubiquitous (things held in your hand)
you could get interesting information to help you notice context-switches from location, mobile phone profile �