Testing the ability of artificial agents to represent human decision-making needs to go beyond checking their ability to reproduce the behavior of corresponding decision-makers and also check their ability to represent their knowledge and cognition.

This article describes the SOSIEL method, which relies on a comprehensive framework of bounded rationality for checking the knowledge, cognition, and behavior of artificial agents, and applies it in testing the ability of two different types of artificial agents to represent the decision-making of participants in a social psychology experiment, called the Voluntary Contributions Mechanism. The article demonstrates that additionally checking the ability of artificial agents to represent the knowledge and cognition of corresponding decision-makers can help identify false positives. The method provides a structured and theoretically-grounded alternative to the Turing Test.