In a famous two-person ultimatum game in which two persons are to split a prize of $10, the first mover gets to make a take-it or leave-it offer to the second mover. If the offer is accepted, they split the money as offered. But if the second mover rejects the offer, neither party gets anything. In thousands of trials around the world, with different stakes, people reject offers of 30% or less.
But if a computer, rather than a human, makes a similar offer, the second mover is more likely to accept it.
Apparently, unfair motive was attributed to human players but not to a computer. So people care about not only outcomes, but what the outcomes mean.
This mistrust of human agents is confirmed by the betrayal game in which the principal, acting through an agent, is willing to accept a lower probability of more favorable outcome when dealing with non-human agents than with human agents. A trustworthy agent would act for the benefit of both resulting in a win-win outcome that is better than the principal acting without the agent. The selfish agent would act for his own benefit at the expense of the principal. In this two-person game, the potential principal would act through the agent only if the agent is at least 50 - 80% trustworthy. But the same potential principal would be willing to take the same risk if the chance of a randomly drawn more favorable outcome is only 30-40%. In other words, the non-human agent only has to be 30-40% trustworthy for the principal to act through the non-human agent. The difference between the two probabilities can be ascribed to the fear of betrayal from even more "trustworthy" people.
The betrayal aversion premium over aversion of pure risk arising from natural events has significant implications in theoretical and real-world problems. In conventional game theory, players are assumed to care about only substantive results. If players also care about human frailty, all the predictions of human behavior become suspect. And in low-trust cultures, many potentially mutually beneficial exchanges simply would not take place.
- Bohnet, I. et al. "Betrayal Aversion on Four Continents." Kenndy School of Government. Working Paper Number:RWP06-005. Submitted: 02/03/2006.
- Lambert, C. "The marketplace of perceptions." Harvard Magazine. March – April 2006.