U.S. Markets closed

Experiment suggests that the best robot bosses could be jerks

Luke Dormehl
Massachusetts Institute of Technology (MIT) has announced a new $1 billion college of computing designed to offer the best possible education to future machine learning A.I. experts.

Provided they don’t put humans out of business altogether, there is a good chance that, at some point in your lifetime, you will find yourself working for a robot boss. But if you think you will have an easier time working for a machine than you do working for your current flesh-and-blood boss, you may be setting yourself up for disappointment.

At least, that is the takeaway from a new piece of research coming out of France’s University of Clermont Auvergne, where investigators have been examining the best way that robots can coax the most productivity out of us as employees. The sad answer? Quite possibly by behaving like jerks.

Their experiment involved the so-called Stroop test, in which different color words appear on a screen, and subjects must identify the color without getting fooled by the word itself. (For example, identifying the word “brown” written in pink as pink, rather than brown.)

For this robot boss variation on the test, participants were first made to have a chat with a robot, which either gave positive (“I think we could become friends”) answers or negative ones (“I do not value friendship”) to questions. The test subjects then took the test. Those who made fewer mistakes, and answered more rapidly, were the folks paired with the meaner of the machines. These subjects performed better than either people paired with friendlier robots or with no robot at all.

As Nicolas Spatola, one of the researchers on the project, told Digital Trends, this effect is not limited to robots; more callous human bosses also prompt similar test results from subjects. However, it can vary according to the difficulty of the tasks, meaning that there’s no one-size-fits-all answer. “There is a theory of challenge and threat from [psychologist Jim] Blascovich that explain this very well,” Spatola said. “The idea is that the presence of others increase our arousal, and according to the difficulty of the task it can be positive or negative.”

Unfortunately, it does mean that whatever more jerk-like traits we see in some human bosses are likely to continue into the age of automation. Heck, it might even be in roboticists’ interest to program them in. Jerkiness could be a feature, rather than a bug. Not that Spatola is necessarily endorsing that idea.

“Even if we show that a bad robot can have a positive effect, we do not know what could happen to individuals if they were monitored by a bad robot during a long period,” Spatola continued. “I’m not sure that it would be good for their well-being.”

A paper describing the work was recently published in the journal Science Robotics.