Greg Detre
Wednesday, October 30, 2002
Ted (first
half), Andrea, Lorin
status of
AI, frames etc. � what supports context?
frames look a lot like a DB � brittle
the statistical ones can�t explain themselves
today � multiple representations
entry
points, activity landscape and coordinating mechanisms
focus on an
office, not a meeting room or whatever
Ted�s office � the point of it is to have as many possible contexts that
he can share with other people (microscope, light table, inventions)
he built: �Room with a view�
wall, low-res projection of an office
showing shelves, photo gallery, desks
so the question that Kirsh is addressing is what should that Room with a
view include?
and how can you make sure that you�ve captured what you need to to
recapture the user�s context?
Ted is making choices for the user � what if you want to add post-its to
a book, or something???
you have to
be clear about what you mean/want by an office
a laptop
lets you take information from one place to the other, but you lose all the
context
unix �
allows you to set up different environments � web server stuff vs email stuff
vs �
they are contexts of work
an application being open is supposed to be a context
many of the
best programmers he�s met don�t use tools (just emacs and a trace for
debugging), because they�re used to having to move environments so often and
losing the tools
operationalising
the concept of work?
do we need an analytical definition of how/what it means to work?
Goms???
�neats vs scruffies� is maybe too broad a categorisation
their goal is not to put things into stacks, but to be able to retrieve
things automatically
so let�s ignore that distinction
why do we want to operationalise the concept of
work?
it has to be general, but also idiosyncratic
you need to come up with the right parameters
that�s ok so long as you don�t leave anything out
I really like his discussion of
operationalisation of work with the intention of keeping it psychologically accessible
and capturing the feel of the office
but what if it�s inherently physical, and just unrealisable
how could you digitalise the way neats + scruffies work?
it has a lot to do with visibility/intrusiveness
learning dimensions (Kagan) of children being more visually/auditorially
attentive
visual is more gestalt, whereas auditory is more sequential
how do you
make the space flexible, so that two people can share the same office by
night/day?
I like the
idea of entry points and coordinative mechanisms
how do you get back to what you�re doing when you�re interrupted?
you look at the papers on your desk etc.
I�m still not sure what activity landscapes are
I�m
starting to feel that the best we can manage (with regard to the portable
office/microvenue) in the medium-term is a laptop and a sheaf of papers
sometimes
we like to take just a laptop and the tools for the task to self-constrain to
the task
part of the problem with a laptop is that it allows you to do so many
things with it
what else
could a laptop do?
it could be more context aware � e.g. you�re in a lecture
its big limitation for me at the moment is its physical environment
multiple screens
scribblable (e.g. on PDFs)
IBM scribblable laptop, MS Scribble (live docs)
most buildings
are built with a particular institution in mind
cellphone messes that up, because it allows you to be a lot of different
things everwhere
(with pervasive computing etc.) the role that you play becomes less a
function of the place you�re in
capture
model
analyse an activity, articulate it in a grammar, impose upon a person in
order to allow them to interact with a computer (to capture the activity in a
computer)
the goal (of context aware computing) is to minimise that imposition on
the user
it�s ok to influence the users� behaviour for a high-integrity reason
(to help them with their task, rather than for my purposes/ease)
Agre
breakfast example
wanted to have a technology to make breakfast for him � all the
different things that it would have to know
Shankian scripts � slice the banana with the knife � what if there�s no
knife? people would use a spoon to cut the banana
context
aware computing is going to have an even harder task in future, because you
won�t just be able to guess at context according to location
3 levels:
architecture
actual place/building
practices
routines
institutions
relationships between humans (persistent structure)
difference between practice + institutions �
institutions include social roles
could you
use statistical methods (e.g. a Markov model) on the vocabulary/formality of
speech at given times as an indication of your context???
like you talk formally with your boss, about your wife with your
friends, about the weather in an elevator
Sandy Pentland � tried to video + speech annotate his life
system tries to predict where he�s likely to go next
searchable audio
social network analysis based on audio � who you�re talking to
radiology
is good for speech recognition because it�s so structured
interestingly: the doctor sounds like the patient when giving a history,
as opposed to the grammatical formalisms when dictating the diagnosis
what does he think of hot-desking???
CHI � computer human interaction
unfortunately,
I think Kirsh underestimates how important little superficial psychological
permanences and habits are to being productive
is there
any evidence to show that hot-desking is unexpectedly
unproductive/disruptive???
somehow, I
was hoping for something a little more concrete from Kirsh about the sort of
representations we might employ for the digital office � if anything, all his
paper serves to do is to show why the digital office cannot be synchronised
with the physical office � the notion of a �portable office� is misleading, if
it implies that you can just create a simulacrum of the physical office � it�s
only right if it means that your one and only office is portable
I don�t
really know what to say about this � what major points does it actually
argue???
it
discusses the �capture� model of how you deal with all this complexity as a
designer by forcing people to bend to the machine�s limitations.
like all of
these monolithic man-made mental structures, how do they adapt themselves???
can the thing build its own frames??? how would you teach it the rules of a new
game, let alone dropping it in a VR simulation and expecting it to act
intelligently or something else really world-upturning�???