Trust Me, I’m a Robot

A recent study suggests methods to increase trust in autonomous machines, improving adoption and cooperation.

Nick Bild
4 years agoRobotics
(📷: C. de Melo and K. Terada)

Autonomous machines are becoming more and more a part of daily life, and that is a trend that, by all appearances, will continue well into the future. Self-driving cars, robots, and personal assistants can make our lives easier in many ways, but will face challenges in adoption and effectiveness unless people are able to trust and cooperate with them.

Researchers from the U.S. Army Research Laboratory have been working to understand the ways that expressions of emotion influence people’s feelings of trust towards, and willingness to cooperate with, autonomous machines.

The team conducted an experiment in which human participants engaged in the iterated prisoner’s dilemma — a sort of game in which two players make a simultaneous decision to either cooperate or defect. If both cooperate, they receive a set payout. If both defect, they both receive a payout less than if they had cooperated. In the case that one cooperates and the other defects, the defector receives the highest possible reward. Players were paired with anonymous counterparts represented by virtual faces able to express emotional states (cooperative or competitive). The counterparts were instructed to follow specific strategies of either generosity or extortion.

The results of the study revealed that both strategy and emotion influenced the participants decisions, with the effect of strategy being stronger than emotion. Participants were found to only be influenced by expressions of emotion when the actions of the counterpart were insufficient to reveal their intentions. In such cases, smiling of the virtual faces after mutual cooperation was found to encourage more cooperation. In contrast, smiling after exploiting the participant would hinder future cooperation. In cases where the counterpart acted competitively, the participants would disregard emotional signals being sent.

I find the results encouraging. They suggest subtle ways in which autonomous machines can better work with humans in more natural-seeming ways. More importantly, the findings show the limitations of the effect — it does not look like a computer generated, pixelated smile will have us bowing to our robot overlords any time soon. Well, at least as long as an AI does not tell us to, anyway.

Nick Bild
R&D, creativity, and building the next big thing you never knew you wanted are my specialties.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles