INTERVIEW:
Question 1
Can you explain what kind of work you're doing with affective computing?
Answer
Affective computing is computing that relates to, arises from or deliberately influences
emotions. Our aim is to give computers the ability to recognize your emotions, to express
emotions, when that's appropriate, to communicate your emotions, and in some cases to
actually have mechanisms inside that are emotional. We also aim to give them the skills of
emotional intelligence, so that they can manage and utilize emotions in a very logical and
intelligent way.
Question 2
How do you do this? Do you use sensors?
Answer
There are many ways to recognize emotions. We are imitating both the way humans recognize
emotions, such as looking at facial expression and looking at gesture and posture,
listening to voice. We're also able to pick up on some other information because of our
development of wearable computers. Wearable computers are worn like clothing. They are
sensors that are put in your shoes, in your jewelry, that you can use as a regular
computer but it can also go around with you so much that it gets to know you. So these are
computers that are with you a lot and have an opportunity to get to know your personal
preferences and values and goals.
Question 3
So if you have one of these computers, you can use it as a personal secretary, and talk to
them and have an assistant, a personal assistant like a human being.
Answer
Yes. The metaphor of assistant is a very interesting one. That's what agents are really
about. It's about building assistants that would be able to be proactive much like a
person is. They don't just sit around and wait for you to tell them to do 1, 2, 3 on this
list, they have some autonomous behaviors and they try to learn your preferences and they
also take feedback from you. So if they do something and you don't like it, and you sort
of frown at it or express displeasure even in a very subtle way, they should be able to
see that and adapt their behavior to you. To do that, though, they have to be able to
recognize things like if you're pleased or displeased, if you're interested or confused.
Question 4
What is the difference between one of these computers and a human person?
Answer
Right now there are a lot of differences between these. The question of what differences
will exist in several decades is largely one the philosophers are battling over right now.
There certainly are some interesting ethical and philosophical questions related to these
machines. If they develop full emotional systems downstream, will we accord them greater
status? Will they threaten our personhood? And these are questions that we are grappling
with at MIT. In fact, there will be a conference this spring addressing the issues of
personhood and human dignity, inviting both theologians and computer scientists to speak
about this topic.
Question 5
To understand the concept can we say that that these computers are like clones of people?
Answer
They are not clones. But what's interesting is even without them having emotions right
now, people already treat computers in basically the same way they treat other people. The
work of Reeves at NASA-Stanford has recently set forth the idea of the media equation,
that people treat computers and even television in many ways like we treat other people.
So even though they don't even have emotions or can't recognize them yet, people treat
computers in much the same way they treat another person. Why express an emotion to an
object that can't recognize it or why type the same thing to it repeatedly when you know
it's not going to do anything different? But the fact is the computer could actually do
something different and it could respond more intelligently to that. And based on the
media equation, we expect that people would like that. So we're giving it a try.
Question 6
But for some people it's almost impossible to reproduce emotions in a machine.
Answer
Psychologists and neuroscientists understand that emotions are both cognitive and
physical. They involve both our thoughts and our physical feelings, our bodily feelings.
For a computer to have emotions, it can have a lot of mechanisms that can imitate our
bodily mechanisms. But it won't have conscious awareness of bodily feelings in the same
way that people do, primarily because it doesn't have the same kind of body that people
do, and it doesn't have the same kind of hormones and the same kind of physical sensors
and sensations that people do. So when we say a machine has emotions, we mean many
different things and none of them are precisely the same as human consciousness of our
emotion feeling. But I think there could be some breakthroughs in giving machines a sort
of awareness of some of their physiological state, and that might be called emotions in
the closest way possible to what humans call emotions.
Question 7
With this affective computer that can understand the mood of someone who is using it,
don't you think that in a certain way the computer will be more and more like a slave of
the human being?
Answer
A lot of people already see computers as tools, and as soon as you humanize that metaphor,
you come up with the slave metaphor. The agent metaphor, which is also a human metaphor
for the computer software, tends to be more the metaphor of servant not slave, more an
autonomous creature that tries to assist you. And as these software and hardware entities
get their own emotional mechanisms, we may see them more like animals in some ways, moving
towards human-like animals. And we might actually be inclined to want to accord them
slightly higher status; it becomes a little bit of a dilemma there. Can you accord any
emotional being the status of a slave or should you elevate it to servant, which certainly
is a much nicer metaphor, since we're all pretty much anti-slavery, but then is there
more? Should they have more rights? And that's certainly an interesting question.
Question 8
In which field do you think that in future this study will be useful?
Answer
Practical things. Some of this sounds like science fiction right now - and some of it is -
but there's also a lot that's science fact. There are also many things that I see coming
out in the near future that are already practical applications. For example, we've built
some eyeglasses that can sense if the muscles around the eye are moving and try to
recognize if it's an expression of confusion - furrowing the brow, or interest - raising
the brow. The information in that signal alone is very interesting to transmit in distance
learning situations, where students are not located in the same room as the lecturer, or
to your computer tutor program that's trying to teach you something. So we've already got
working prototypes of something that's useful right now.
Question 9
Are these affective computers very expensive, or do you think that in the future everybody
will have an affective computer on their desk?
Answer
I think the box on the desk is going to change dramatically, and we'll see more and more
things like pencils and pens and your shoes and various products acquiring computational
abilities without looking anything like present-day computers. Some of them will have no
emotions and some of them will have very sophisticated sets of emotions. We'll find a
complete range, much like you find a complete range of animals, from very simple insects
that may only exhibit a flight or fight kind of behavior - you might say they have just
the simple emotion of fear - on up to human beings that have, of course, the most
sophisticated repertoire of emotions. So we will find a whole range of emotional abilities
and different things and not just the box on the desk.
Question 10
Do you think that in the future the computer will be like a human being?
Answer
No. I don't think computers will be like human beings. Many computer scientists these days
are saying that computers will evolve to be much smarter than people. Certainly they will
have much better abilities than us in some areas, but I really think that computers and
people are very different and will remain quite different.
|
