Affective Computing: Teaching Machines About Emotion

jbc writes "The L.A. Times is running a story about affective computing, a field in which researchers are programming computers to recognize human emotions through the use of such clues as facial expression, vocal tone, and blood pressure. Some hail it as the dawn of a new era in super-useful machines, while others warn about invasions of privacy."

Comments

At best these will have very poor reliability

[Anonymous Coward]

The types and amounts of emotions got from automated physical measures will be a small primitive and not to reliable set. If you think you'll get much more then think again. Now out dated behavioural psychology failed/was very limited in this approach. Also, check out things like how specific lie detectors are.

According to many clinical psychologists, when it comes to predicting what a patient will do next, (which of course is also based on feelings), the most reliable way to find out is just to ask them.

Emotion is thought.

[kmellis]

From the article:

Movellan is part of a growing network of scientists working to disprove long-held assumptions that computers are, by nature, logical geniuses but emotional dunces. The ability to interpret markers for emotion--facial expressions, vocal tones and metabolic responses such as blood pressure--may seem like crude first steps. Yet experts see machine intelligence, unswayed by human frailty and bias, as an eventual advantage. They envision machines that know us better than we know ourselves.

The idea that consciousness could occur outside the context of emotion is a pernicious misconception. It arose from the combination of a greatly oversimplified view of thought and the legacy of dualism.

It is true that what we experience as "emotion" is a subtly different kind of cognition than what we experience as deliberate thought. It's more fundamental, and more closely tied to other physical systems. So I do agree that it makes a certain sort of sense to distinguish "thought" and "emotion". Ultimately, however, both are manifestations of the same fundamental brain activity. They are deeply related and are not in opposition.

We've been spectacularly bad at analysis of our own consciousness. History has shown that much of what we don't notice and so take completely for granted are fundamental and extremely difficult problems; while what we are very aware of and have concentrated upon have proven to be trivial. The predicate calculus, in this context, is trivial.

I've long railed against the cliche of the "unfeeling" thinking machine/being one sees in popular science fiction. Neither Spock nor Data would be able to carry on a meaningful conversation if their thought didn't exist within the context of emotion. The idea that a thinking machine could imitate human consciousness without including human emotion is absurd if examined carefully.

Be that as it may, "affective computing" is only a very minor addition to computing in the context of AI. It's just another form of data acquisition, albeit one that would no doubt be very useful for an AI. None of this stuff we hear about is even remotely close to actual AI; at best it's just "smarter" computing. Real AI will only be achieved when we are able to build (or more properly, "grow") very high-level complex adaptive systems aimed at complex human interaction.

Thanks for Doing LA Times' Job For Them

[GuyMannDude]

I'd suggest reading AffectiveComputing by Rosalind Picard from MIT Press, her homepage is here [mit.edu] and interview on First Monday [firstmonday.dk] and the MIT homepage at MIT [mit.edu]

Thanks for posting this. For the LA Times' article, one gets the feeling that Movellan is leading a one-man renaissance in AI. Like most articles about far-future technologies, the article is heavy on the "gee-whiz" and "what will they think of next?" stuff and light on any sort of in-depth examination of the issues involved. First, I don't understand why the media (newspapers especially) don't take the time to do a thoughtful, in-depth story about non-time-critical issues like Affective Computing. Secondly, I wish that if they were going to do a half-assed job of it, they would at least cite other, more detailed sources of information so interested readers could learn more on their own. Yeah, I suppose someone can do a web search to find this out. And thank god for slashdot where the readers usually know more about the subject matter than the article authors. But it's common curtosy to cite important people in a scientific field. At least it is when writing a scientific paper -- why should the mass media be exempt from this little niceity? Suppose you were a researcher at MIT's program in this field and saw this article. Wouldn't you be kind of pissed off? The LA Times could have replaced that paragraph about the Golem with a paragraph about the MIT program.

I'm troubled by the slipshod coverage that science and technology gets in the mass media. Do the newspaper authors think we don't care to know the details? Do they themselves not care?

GMD