Abstract
Quantification of human behavior in a social context may lead to discovery of subtle behavioral variations in a population that can be used to improve classification and screening for psychiatric disorders, and provide more accurate targeting in the development of interventions and biomedical treatments. However, it is difficult to study social interaction in a controlled, reproducible environment, as well as analyze the resulting behavior. In this research, we describe an experimental framework that utilizes a game of iterated Rock-Scissors-Paper played against an artificial intelligence agent, and a behavioral hypothesis of rule-switching to motivate analytical methods, that will extract behavioral features with potential clinical relevance from game data. Clinical relevance Presently, psychiatric diagnosis involves subjective assessment of symptoms. Finding quantifiable, observable behavior that displays a spectrum in a population may be useful to differentiate and diagnose psychiatric illness.
All Related Versions

This publication has 66 references indexed in Scilit: