Skip to main content

Is That a Renoir, or an R2D2?

Humans are wary of computer-generated art, but new research suggests our views on the subject are malleable.

As robots take over more and more jobs that were once performed by human beings, one of the few occupations that seems safe is that of the artist. New research offers some comfort for creative types, finding the public takes a dim view of computer-generated art.

But don't rest too easy. It also finds our bias in favor of human-produced work declines when we have a chance to observe robot artists in their natural environment—and especially if we begin to anthropomorphize them.

If Wall-E starts emulating Warhol, watch out.

A research team led by University of London psychologist Rebecca Chamberlain and Caitlin Mullin of the MIT Computer Science and Artificial Intelligence Laboratory describes two studies conducted at the University of Leuven in Belgium. The first featured 65 people, 20 of whom were knowledgeable about art.

Participants viewed 60 images of artworks, half of which were generated by computers, and rated the attractiveness of each on a one-to-seven scale. They also guessed whether it was created by a human or a computer. The pieces were "broadly matched" for both content and mode of production.

"Images that were categorized as computer-generated were rated as visually less pleasing," the researchers report in the journal Psychology of Aesthetics, Creativity, and the Arts. This held true whether participants appraised their quality before or after classifying its likely genesis. (They were pretty bad at guessing: The mean accuracy rate was only 52.5 percent.)

We may enjoy Robocop, but we're not quite ready to embrace Roborembrandt.

The second study involved three sets of participants. The first featured 145 people who were approached at a Brussels art museum. They viewed both a series of drawings and the non-human artists that created them.

The robots all had "a left-handed planar robotic arm with a black Bro pen," and were "positioned around a chair where the subject was seated, much like a real drawing class," the researchers write. After checking them out, visitors reported their impressions of the robots, and evaluated their drawings.

The second group, 97 participants, viewed only the drawings (all facial portraits).They were informed that the works were created by robots, and read about the process by which they were created. Members of the third group, 107 participants, looked at the drawings and received no information on their source.

The results: The first group—which actually saw the robot artists—rated the drawings the highest in terms of aesthetic value. People who simply read about their non-human origin gave the lowest ratings, while those who received no information fell in between.

While the robots "were not perceived as being human or lifelike," some members of the first group "noted that the instillation did have the feel of a drawing class, with intelligent agents systematically observing and recording," the researchers report. Engaging in this sort of anthropomorphism "correlated positively with aesthetic response."

The results suggest computer-generated art could gradually gain acceptance—but only if and when people come to believe that the robots are genuine creative agents, rather than machines imitating human activity. We may enjoy Robocop, but we're not quite ready to embrace Roborembrandt.