Trust and Trustworthiness : Experiments with Artificial Intelligence (AI) Agents

The trust game, a simple two-person economic exchange, has been extensively used as experimental measures for trust and trustworthiness of individuals. Here, we develop deep neural network-based artificial intelligence (AI) agents to participate a series of experiments based upon the trust game. These artificial agents are trained by playing with one another without any prior knowledge or assumption regarding trust. We find that artificial agents can mimic trusting and trustworthy behaviors and produce actions that are close to decisions by human subjects. Conditions that influence levels of trust and trustworthiness by AIs are further explored. This study offers evidence that trust can emerge purely from an interactive learning process and demonstrates the possibility to build AI-integrated decision support systems capable of leveraging social behaviors to achieve better outcomes collectively