14 September 2012

A Robot Can Help Teach Us

An unusual new study of college students’ interactions with a robot has shed light on why we intuitively trust some people and distrust others. While many people assume that behaviors like avoiding eye contact and fidgeting are signals that a person is being dishonest, scientists have found that no single gesture or expression consistently predicts trustworthiness. But researchers from Northeastern University, the Massachusetts Institute of Technology and Cornell recently identified four distinct behaviors that, together, appear to warn our brains that a person can’t be trusted. The research could one day be used to develop computer programs that can rapidly assess behavior in airports or elsewhere to flag security risks. In the first experiment, 86 undergraduates from Northeastern were given five minutes to get to know a fellow student they hadn’t met before. Half the pairs met face to face; the other half interacted online by instant message.


Then the students were asked to play a game in which all the players got four tokens and the chance to win money. A token was worth $1 if a player kept it for himself or $2 when he gave it to his partner. Players could win $4 each if both partners kept their tokens, but if they worked together and traded all four tokens, then each partner could win $8. But the biggest gain — $12 — came from cheating a partner out of his tokens and not giving any in return. Over all, only about 1 in 5 people (22 percent) were completely trustworthy and cooperative, giving away all their tokens so that each partner could win $8. Thirteen percent were untrustworthy, keeping all or most of their tokens. The remaining 65 percent were somewhat cooperative, giving away two or three tokens but also holding one or two back for security. Both groups demonstrated the same level of cooperation. Whether the students met face to face or online didn’t change their decisions about how many tokens to give away or keep. But students who met in person were far better at predicting the trustworthiness of the partner; that suggested they were relying on visual cues.

More information: