Deception Detection by Humans in the Iterated Prisoners Dilemma

For humans, lying is a complex thing to detect, both when lying and when being lied to. While telling a stone cold lie might be difficult for humans, it is relatively easy for a computer to do so. People pick up on lying peers by some involuntary telltale signs and emotions that are easy to mask on a computer image as it does not suffer from such human traits. Thus when playing a game, actions can speak louder than words; people might tell one thing, but do another. This study examines whether humans can pick up on an Embodied Conversational Agent (ECA) portraying emotions belonging to a certain strategy, when actually playing the opposite strategy and if they act accordingly. This was done by having participants play a game of the Iterated Prisoners Dilemma (IPD) against an ECA opponent. The strategy of the ECA and the fact whether the ECA displayed conflicting emotions or emotions that are consistent with the played strategy were varied between participants. Afterwards participants were asked what they thought of the ECA they had just played against through a short questionnaire. An online game was developed and 42 persons participated. Out of those 42 participants, 39 completed the questionnaire at the end. The other three participants ended the experiment after having played the initial game. The results showed no significant differences between the researched groups, both in terms of emotions (the action taken) and in terms of cognition (the questionnaire). Some filtration and interpretation of the data, combined with results from previous research into the subject, might warrant further research, as the subject itself is still very much topical in today’s science.