Cassell is an associate professor at the MIT Media Lab
where she directs the Gesture and Narrative Language
Research Group. She holds a master's degree in Literature
from the UniversitÚ de Besanšon (France), a master's
degree in Linguistics from the University of Edinburgh
(Scotland), and a dual Ph.D. from the University of
Chicago (U.S.) in Psychology and in Linguistics.
builds systems that look like humans and that have some
of the same kinds of social and communicative competencies
that we do. Anthropomorphic systems such as these can
lower the bar to computer use, allowing access to computers
to those who can't type or read, for example. Increasingly
Cassell is turning her attention to the role that technologies
such as these may play in children's lives.
technologies such as a virtual storytelling peer have
the potential to encourage children in creative, empowered
and independent learning. They can also demonstrate
ways for interactive technology to live away from the
desktop, supporting children's full-bodied, collaborative,
social play-based learning. Justine Cassell is co-editor
of "From Barbie to Mortal Kombat: Gender and Computer
Games" (MIT Press, 1998), and of "Embodied Conversational
Agents" (MIT Press, 2000), and she has published extensively
in journals as diverse as Poetics Today, Pragmatics
and Cognition and Computer Graphics.
scientist's answers are available below.Please see our resources
page for the scientists home page and other related infomation.
was interested in your part of the show. If the characters
can be used to elicit a child's response, can this mechanism
be used to help to change a person's behavior? I am
a health educator and I am interested in VR in my field.
How can I get more information? Thanks for any help
you can give me.
Dear Pat, We do believe that these characters may
also have their uses in behaviour change. One of my
graduate students is currently writing his Ph.D. on
exactly this question, in fact. We know that simple
interaction with a health advisor can help adherence
to behavioural change programs. And so, he has designed
a health advisor character that can try to help people
start on, and continue with, an exercise program. The
study that he is running looks at whether it is more
effective to have a character, or text only, and in
each of those conditions, whether social talk helps
the interaction. You can look at his work on our Web
page, under the rubric: "social language".
you considered the possibility of using archives from
various chat rooms to build a database of social interaction,
to teach your systems what is acceptable and what is
rejected in conversation?
We ourselves have not used archives from chat rooms,
but some designers of 'bot systems (virtual characters
that "live" in chat rooms) have indeed followed that
strategy. In fact, in our domain, learning algorithms
(that is, using databases of previous interaction to
"teach" the agent) are quite difficult to implement,
because there are so so many different aspects to the
interaction that the character must learn: body language,
prosody, content of language . . .
am hearing impaired. I read body language and lips extremely
well to compensate for what I cannot clearly hear. (My
dad says I can "hear" as long as I don't actually need
to hear.) I hate watching cartoons because I cannot
read their unrealistic lip movement or their body language.
What do you plan to do with your program for people
like me who do not know sign language but need the added
real world stimulus of lip movement to be clear on what
is being said? At this point, your program is just another
cartoon character to me that I cannot hear or clearly
We ourselves have not carried out work on accurate
reproduction of visemes (that is, the different shapes
of the lips that correspond to the phonemes of sound),
but wonderful work on this topic does exist. For example,
Dominic Massaro and his research group have built very
accurate talking heads to convey spoken language more
clearly to the hearing impaired, and to help hearing
impaired children learn to read lips.
Justine, I was wondering if you had ever considered
using Sam as a program to help children who have difficulty
with social interactions? My son is diagnosed with Asperger's
Syndrome, a mild form of autism. Some of his difficulty
is in understanding the rules of social interactions.
It appears to me that Sam might be a very useful tool
to help children develop that sense. Have you ever thought
of using the Sam program in that capacity? Thanks.
question is one that has interested me for a long time,
but that has not been addressed by the research community,
to my knowledge. I've heard that autistic children respond
better to representations of humans that are computer-generated
than to real humans, which would lead us to think that
virtual children of this sort would make good playmates,
and potentially peer instructors, both of social skills
and other kinds of skills. I'd love to have my work
used for this purpose, but I don't have the training
to carry out the work myself with autistic children.
I hope to see this kind of work carried out in the future.