Friday, May 11, 2007

Anthropomorphism and robotics

Spent the last two days at a really interesting workshop: Practices of Anthropomorphism: from Automata to Robotics. The best thing about the meeting was the mix of disciplines, including anthropology, psychology, art and robotics. Terrific. The key to creativity, IMHO, is working with people outside your own discipline. It can be tough, mind you, outside the comfort zone of familiar ideas and frameworks. But that's precisely the point. Recall Koestler's brilliant "The Act of Creation" - creativity happens when two previously disconnected ideas intersect, meaningfully. (Or, humour, if the intersection is absurd.)

So, what were the creative outcomes of the workshop? Well, there were plenty of "ah ha!" moments (as well of lots of "ha ha"). For me the big insight was the realisation (obvious I guess to the anthropologists) that human beings are so utterly pre-disposed, hard-wired even, to anthropomorphise. Whether we like it or not we are, it seems, anthropomorphiliacs with a compulsion to attribute anthropic qualities to and develop emotional attachments with animals or artefacts, almost anything from the everyday to the exotic, the banal to the sublime. This is very interesting to a roboticist on a number of levels. For example, even simple robots are imbued with characteristics they don't have (cute, inquisitive, happy, ill) and at the other end of the robot spectrum humans will anthropomorphise and thereby compensate for the shortcomings of humanoid or android robots. In other words robot builders don't need to worry about making android robots into perfect artificial humans - cartoon robots will do - like NEC's PaPeRo shown here (thus also neatly avoiding the Uncanny Valley). You may think this is just another fraud, like de Vaucanson's defacating duck. And, in a way, it is. But as long as we, the roboticists, are completely honest and transparent about the real capabilities of our robots: what they can but especially what they cannot do, then it is a fraudulent contract that humans and robots can willingly and beneficially submit to.

This brings me to another question. Why are we so fascinated by robots? I think the answer lies in another surprising innate ability of humans. That is our ability to tell the difference between animate and inanimate. I think we love robots because they behave as if they are animate, yet we know them to be inanimate artefacts. We are, I believe, delighted by this deception.

A final thought: if, as Gabriella Airenti brilliantly argued at the workshop, anthropomorphism is an (inevitable) consequence of imitation and theory of mind, then it's surely not inconceivable that future intelligent robots might also develop this tendency. Except that, for them, it would be robomorphism.

1 comment:

  1. Recently I read that anthropomorphism is generally taken as being a negative thing. I hadn't previously taken it to be either good or bad in itself. Maybe the effects of doing this may result in things going wrong.

    There is also a lot of difference between simply believing something is human and having some hard wired responses to a thing (robot/media/animal) because we are human. One is a naive approach and the other is just who we are, naturally social creatures.

    You may have seen the work of Byron Reeves and Clifford Nass before. Well worth a read.