Plausible Deniability and Cooperation in Trust Games

What motivates agents to choose pro-social but dominated actions in principal-agent interactions like the trust game? We investigate this by exploring the role higher-order beliefs about payoffs play in an incentivized laboratory experiment. We consider a variety of ways of distributing higher order information about payoffs, including an asymmetrical distribution that generates "plausible deniability": one agent (B) knows the other (A) doesn't know that B knows how A's payoffs are impacted by B's actions. Agents, in turn, exploit this: otherwise trustworthy types are tempted into defecting when they have plausible deniability.