The ghost in the machine

A recent edition of The New Scientist carried an intriguing article about an experiment in social psychology.

A recent edition of The New Scientist carried an intriguing article about an experiment in social psychology. Two Harvard researchers, interested in human altruism, asked volunteers to play a gambling game in which altruism paid off, but only when it was reciprocated. The group was divided into two; half of the subjects were asked to make their choices under the benign gaze of a robot called Kismet, specifically designed to be able to mimic human expressions.

No one could mistake Kismet for a human being, but no one would miss the fact that "she" had a face either - and when she was observing, the participants donated 30 per cent more than when she wasn't, even though Kismet wasn't ostensibly part of the experiment at all. Subconsciously, the Kismet players behaved as if they were being watched by someone, and so they behaved better. No research was done into the incidence of nose-picking or crotch-scratching, but my hypothesis would be that they fell sharply in Kismet's presence.

The experiment appeared to confirm that at certain subconscious levels we seem to be robots too. Punch in the right stimulus - a pair of dots over a gently upturned curve - and we see a face. Add a little more detail (surprisingly little) and we imagine a psychology behind it - a conceptual space into which we are likely to project our prejudices and attitudes.

This hard-wired tendency to respond to anything remotely human-like as if it were human is what makes films such as Robots - the latest digital animation offering - possible in the first place. Nobody in their right mind would pitch a heart-warming tale about an outboard motor, an ashtray and a fire hydrant, but tweak the hardware into a simulation of a face, and we'll happily go along with the idea that there are ghosts in the machine.

Well, not all that happily in this particular case, because Robots is so mechanical in its operations that no amount of visual dazzle can overcome its inherent deadness. A software programme could have developed the plot (and its version might have contained fewer logical contradictions). It doesn't even do the audience the courtesy of concealing the fact that it's being treated as a programmable machine - press button A for laughter and button B for pathos. More serious, Robots doesn't understand that the most compelling movie robots touch us on precisely this nerve - the uncertainty about what separates an artificial intelligence from its flawed human counterpart. At their best, they're not cuddly and lovable; they're philosophically unnerving.

A Japanese roboticist called Masahiro Mori helped to explain why this is. He plotted the empathy and warmth that people felt against the degree to which a robot looked like a human and discovered something interesting. Generally, the more human the robot looked, the more empathetic the response. But the curve included a deep dip when the sense of connection suddenly flipped to aversion. Miro called it the "uncanny valley", the point when the robot was human-like enough to trip the switches of fellow-feeling but was still robot-like enough to trigger subliminal alarms.

The truly interesting robot moments in movies take place in Uncanny Valley. Cute robots - C3PO and R2D2 or the mechanical chimp in Short Circuit - all live down-slope of that canyon, which is why our pleasure in them is essentially frivolous. They are cartoons.

But think of the one truly memorable scene in I, Robot: when a single set of eyes suddenly flickers with independent life as you track along a row of identical metal faces. That's deep in the shadows of Uncanny Valley - an object that is manifestly inanimate suddenly displays one of the most and fine-grained markers of consciousness.

And think too of the opening 50 minutes of AI. There's a scene between David, the child robot, and his mother/owner that is piercingly creepy about the distinctions between real and simulated feeling. In one sense, David doesn't inhabit Uncanny Valley at all - in terms of appearance, he's completely indistinguishable from a real little boy. But that's his address all the same, because he occupies that deeply unnerving transitional space between mere mechanism and us.

These moments and others like them aren't unsettling because they suggest that robots may one day displace humans (the basic plot of so much robot horror) but because they raise the awful possibility that we may just be a kind of robot ourselves. It isn't a sheer cliff that divides us from the bottom of Uncanny Valley - it's a slope, and we may be much farther down it than our vanity will allow us to admit.