WordSpy: Affective Computing

Eric Tiedemann sent me to WordSpy, the "website devoted to lexpionage". The term affective computing reminds me of the book of that title I read a few years ago, by Rosalind Picard. The basic theme of the book is artificial emotional intelligence. My thinking is that emotional intelligence, as a component of artificial intelligence, is a worthy and attainable goal. I see this as a branch of natural language understanding and synthesis; the language of emotion communicates very compactly about things that are very important.

I don't think that artificial emotion is either possible or desirable. One can endow artificial agents with a utility function, and from that utility function they may derive further goals and preferences; but the overall mission will ultimately be externally directed. From Robert Pirsig's Zen and the Art of Motorcycle Maintenance, here's a quote on Quality:

In my mind now is an image of a huge, long railroad train, one of those 120-boxcar jobs that cross the prairies all the time with lumber and vegetables going east and with automobiles and other manufactured goods going west. I want to call this railroad train "knowledge" and subdivide in into two parts: Classic Knowledge and Romantic Knowledge.

In terms of the analogy, Classic Knowledge, the knowledge taught by the Church of Reason, is the engine and all the boxcars. All of them and everything that's in them. If you subdivide the train into parts you will find no Romantic Knowledge anywhere. And unless you're careful it's easy to make the presumption that's all the train there is. This isn't because Romantic Knowledge is nonexistent or even unimportant. It's just that so far the definition of the train is static and purposeless. This was what I was trying to get at back in South Dakota when I talked about two whole dimensions of existence. It's two whole ways of looking at the train.

Romantic Quality, in terms of this analogy, isn't any "part" of the train. It's the leading edge of the engine, a two-dimensional surface of no real significance unless you understand that the train isn't a static entity at all. A train really isn't a train if it can't go anywhere. In the process of examining the train and subdividing it into parts we've inadvertently stopped it, so that it really isn't a train we are examining. That's why we get stuck.

The real train of knowledge isn't a static entity that can be stopped and subdivided. It's always going somewhere. On a track called Quality. And that engine and all those 120 boxcars are never going anywhere except where the track of Quality takes them; and romantic Quality, the leading edge of the engine, takes them along that track.

Romantic reality is the cutting edge of experience. It's the leading edge of the train of knowledge that keeps the whole train on the track. Traditional knowledge is only the collective memory of where that leading edge has been. At the leading edge there are no subjects, no objects, only the track of Quality ahead, and if you have no formal way of evaluating, no way of acknowledging this Quality, then the entire train has no way of knowing where to go. You don't have pure reason...you have pure confusion. The leading edge is where absolutely all the action is. The leading edge contains all the infinite possibilities of the future. It contains all the history of the past. Where else could they be contained?

The past cannot remember the past. The future can't generate the future. The cutting edge of this instant right here and now is always nothing less than the totality of everything there is.

Value, the leading edge of reality, is no longer an irrelevant offshoot of structure. Value is the predecessor of structure. It's the preintellectual awareness that gives rise to it. Our structured reality is preselected on the basis of value, and really to understand structured reality requires an understanding of the value source from which it's derived.

One's rational understanding of a motorcycle is therefore modified from minute to minute as one works on it and sees that a new and different rational understanding has more Quality. One doesn't cling to old sticky ideas because one has an immediate rational basis for rejecting them. Reality isn't static anymore. It's not a set of ideas you have to either fight or resign yourself to. It's made up, in part, of ideas that are expected to grow as you grow, and as we all grow, century after century. With Quality as a central undefined term, reality is, in its essential nature, not static but dynamic. And when you really understand dynamic reality you never get stuck. It has forms but the forms are capable of change.

To put it in more concrete terms: If you want to build a factory, or fix a motorcycle, or set a nation right without getting stuck, then classical, structured, dualistic subject-object knowledge, although necessary, isn't enough. You have to have some feeling for the quality of the work. You have to have a sense of what's good. That is what carries you forward. This sense isn't just something you're born with, although you are born with it. It's also something you can develop. It's not just "intuition," not just unexplainable "skill" or "talent." It's the direct result of contact with basic reality, Quality, which dualistic reason has in the past tended to conceal.

Now to put it briefly, artificial agents cannot directly perceive Romantic Quality. But if they understand the language of emotion, they will have a hotline to Quality through the people around them, and thus they can better serve.

Rosalind Picard directs Affective Computing at the MIT Media Lab. Their many very interesting projects include efforts to synthesize emotions. My sense is that they are using emotion as a shorthand computational heuristic, which I think is legitimate and likely to be very useful. But as I've made clear above, I think there's much more to it than that.