Towards a "Perfect" Digital Assistant. Is a change in the personal computing paradigm in the offing?

mc schraefel

(Southampton University, CS)

In 1992, the "personal digital assistant" or PDA
was introduced, embodied as the first Apple
Newton. A variety of PDA tools have followed from
the Palm to the increasingly ubiquitous smart
phone, each bringing together a set of
applications: calendar, note pad, to do's and
address book - tools that work adequately well
for the narrow percentage of people who are
organized, schedule things anyway and can
tolerate filling in forms.

The precursor of the PDA was the more visionary
1987 concept of the knowledge navigator: an
imaginary Apple vision of what a PDA might be
when it grew up: something that would actively
blend awareness of personal and public data to
provide support for the human operator. It could,
for instance, present comparison of stats from
discrete papers, handle phone calls in the
background while other activities were taking
place in the foreground, such as preparing notes
for a class. While the current PDA is focused on
organizational tools that require constant and
manual attention, the knowledge navigator was
focused on supportive, dynamic interaction based
on existing and new public and private
information blending.

Recently we've been looking at what it might take
to bring the current state of information
management tools away from their manual
application/form filling capacity towards the
vision of the knowledge navigator, or what we've
been calling the "perfect digital assistant."

We are exploring the first part of this process
towards the new PDA in a project called Jourknow
between MIT and Southampton. Here, we're
investigating different desktop models for data
representation and storage. Fundamentally we have
also been looking at new ways to capture the
structure of information but in a form free way.
We have then been exploring how this information
can be enriched for retrieval by automatically
associating it with what we can know about the
context of the information at the time of capture
- what we were doing; who we might be with; where
we were; any applications/documents open. This
project has very much captured on improving
personal information management by improving
capture and retrieval contexts. While we hope the
approach is better than current ap/form PIM
models, it is not the knowledge navigator. The
active assistance component is missing.

In a new project, Idoru, we have just begun to
look at how we might blend public data sources
like rss feeds and other mine-able Web resources,
with ubiquitous sensor or data feeds, and these
with our personal data in order to begin to see
how this might be purposeable for active support:
if our knowledge navigator is watching sources of
data for things we care about, and also knows our
schedule, can it find a hole in our schedule to
alert us to an opportunity for a new gallery
opening for instance, and then book the tickets
on approval? Could it likewise, knowing we've
been pulling all-nighters for a CHI deadline,
recommend something we should have for dinner to
help replenish our health? We have developed an
early prototype, AtomsMasher, to explore how we
can push existing RSS technologies to support
such interactions.

In this talk, i will go over the work informing
these projects and our findings to date, with an
eye towards exploring with you shared interests
and possible collaborations towards a perfect
digital assistant.
Thursday 22nd November 2007, 14:00
Robert Recorde Room
Department of Computer Science