Seeing Minds in Machines: Mind Perception and Anthropomorphism

Whether a robot seems to have a mind depends less on what is inside it than on what is going on inside us, and psychology now has a fairly precise account of which buttons get pressed.

Triangles with intentions

In 1944 Fritz Heider and Marianne Simmel showed people a short animated film of two triangles and a circle moving around a box. Nothing in the film was alive, yet viewers readily described it as a story, with characters who chased, bullied, hid and rescued one another. The experiment has become a parable for social robotics. Humans do not wait for proof before attributing minds; they reach for the social interpretation first and ask questions later, if ever.

Robots are, if anything, far richer stimuli than Heider and Simmel's shapes. They move with apparent purpose, turn their heads, speak and respond. So the interesting question is not why people anthropomorphise robots, but why they do it so unevenly, and what exactly they are attributing when they do.

Two dimensions, not one

A landmark answer to the second question came in a short paper in Science by Heather Gray, Kurt Gray and Daniel Wegner in 2007. Using an online survey completed more than two thousand times, they asked people to compare 13 characters on a range of mental capacities. The characters included adults, children, an infant, a foetus, a person in a persistent vegetative state, a dog, a frog, a chimpanzee, a dead woman, God and Kismet, the expressive robot head developed at MIT.

When the ratings were analysed, they did not line up on a single scale from less mind to more mind. They fell into two clusters. The first, which the authors called Experience, covered the capacity to feel things such as hunger, fear, pain, pleasure and joy. The second, Agency, covered capacities such as self-control, planning, memory, communication and moral judgement. God, for instance, was seen as high in Agency but low in Experience. The robot was credited with a little Agency and almost no Experience.

The two dimensions turned out to predict different moral intuitions. Judgements about who deserved punishment for causing harm tracked Agency, whereas reluctance to harm a character tracked Experience. In philosophical terms, Agency maps on to being a moral agent, responsible for one's actions, and Experience maps on to being a moral patient, someone to whom things can be done. That distinction turns out to be enormously useful for thinking about robots, which are often granted a little of the first and almost none of the second.

Why some people see more

The first question, why anthropomorphism varies, was tackled the same year by Nicholas Epley, Adam Waytz and John Cacioppo in their three-factor theory in Psychological Review. They proposed that people humanise non-human things more when knowledge about humans is readily available and seems to apply, which they called elicited agent knowledge; when they are motivated to explain and predict something's behaviour, which they called effectance motivation; and when they lack social connection, which they called sociality motivation.

Each factor has a direct counterpart in robot design. Humanlike faces, voices and movements make human knowledge easy to apply. Unpredictable or opaque behaviour invites people to explain it by imagining intentions. And loneliness, the factor with the sharpest ethical edge, makes people more inclined to see minds where there are none. In a follow-up set of studies, Epley and colleagues found that both chronically lonely people and people induced to feel lonely were more likely to attribute humanlike mental states to gadgets and pets. Companion robots aimed at isolated older adults are, in effect, designed for the audience most primed to see a mind.

In a later review, Waytz, Gray, Epley and Wegner pulled these threads together. Causes of mind perception, they argued, come from both sides: the perceiver's needs, such as the desire for connection, and the target's features, such as similarity to oneself. The consequences are just as important, because granting something a mind gives it a claim on our moral concern and makes its actions feel meaningful.

The trouble with feeling machines

There is a twist. In 2012 Kurt Gray and Daniel Wegner reported experiments suggesting that what makes humanlike machines unsettling is not appearance as such but the attribution of Experience. A machine described as able to feel was rated as more unnerving than one described as able merely to act, even without a humanlike body. People seem comfortable with robots that plan and calculate, and uneasy with robots that might suffer.

For designers, that is a striking constraint. The features that make a robot feel present and relatable push it towards the Experience dimension, which is also where discomfort and moral obligation live. Mind perception is not a dial you can turn up for engagement and expect nothing else to change. Every increment brings questions about what the machine is owed, and about what its makers are implicitly claiming.

References

1. Heider, F. and Simmel, M. (1944). An experimental study of apparent behavior. The American Journal of Psychology, 57(2), 243-259. Link
2. Gray, H. M., Gray, K. and Wegner, D. M. (2007). Dimensions of mind perception. Science, 315(5812), 619. Link
3. Epley, N., Waytz, A. and Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864-886. Link
4. Epley, N., Akalis, S., Waytz, A. and Cacioppo, J. T. (2008). Creating social connection through inferential reproduction: Loneliness and perceived agency in gadgets, gods, and greyhounds. Psychological Science, 19(2), 114-120. Link
5. Waytz, A., Gray, K., Epley, N. and Wegner, D. M. (2010). Causes and consequences of mind perception. Trends in Cognitive Sciences, 14(8), 383-388. Link
6. Gray, K. and Wegner, D. M. (2012). Feeling robots and human zombies: Mind perception and the uncanny valley. Cognition, 125(1), 125-130. Link

Feature drafted with AI from the published research listed above, for social-robotics.com.