Android robots are entering human social life and the people that are interacting with them are unlikely to be technical experts and more likely to approach these interactions casually and intuitively. Because of this, it is becoming more important to study robot-human interaction. There has been shown dislike towards robots because of their imperfect “humanness”. This has been termed the “Uncanny Valley (UV)” in which imperfect human-likeness provokes dislike.
The first study examined these dislikes and effects with a sampled of 80 robot faces. They rook the 80 robot faces off of a Google search and found participants via a crowdsourcing platform. The sample was generally younger, more educated and lower income, but more demographically diverse than typical research samples. The study was to determine if human reactions to robots exhibit the UV effect and “if so, to determine the degree to which it actually influences humans willingness to trust a robot as a social partner”. They judged humans reactions to the robots’ faces for likeability, and then tested if this UV effect affects social decision making. The participants viewed a page of the 80 faces one at a time and were asked to use a rating scale based on the questions “how mechanical does this robot look?” and “how human does this robot face look?” And finally “ how much positive or negative emotion is this robot face showing?”.
The second part dealt more directly with trustworthiness of robot faces. Instead of rating faces on the 0-100 scale, they used the financial game-theory method to see their willingness to trust the robots in a game with financial consequences. They were asked to allocate an amount to entrust to each robot. As observed for likability, robots perceived as showing more positive emotion tended to elicit more trust.
The second experiment took digitally morphed images of human faces with robot faces to create varying degrees of human-looking robots that don’t actually exist but the participants believed existed in real life. They were asked if they would interact with certain robots based on the faces and if they would enjoy the interactions. The most fully human face was considered significantly more likable than all the others.
In conclusion, “to a point, likability increased with increasing human-resemblance beyond the nearly neutral reactions elicited by the most mechanical robots”. But as faces became more human than mechanical, they began to be perceived as unlikable. Finally, as faces became nearly human, likability was sharply positive again. This steep final increase is an important feature in and “suggests that although the most human-like robots may be more likable than reliably perfect human likenesses, they may occupy a precarious position at which small faults in their humanness might send the social interaction tumbling”.