Learning Human Decision Models


Speaker


Abstract

Human behavior in many decision making tasks are suboptimal when performance is evaluated against an ideal model that knows the structure of the environment. I will show that optimal decision making models that try to learn the structure of the environment show the same kinds of suboptimality.  In particular, models that try to learn environmental volatility (non-stationarity) and learn the value of outcomes capture key aspects of human behavior, like limited memory and repeatedly choosing options believed less valuable.  I will also describe our recent experimental research testing the idea that apparently suboptimal human decisions are a byproduct of a rational and optimal attempt to learn the structure of the environment. 
 
Contact information:
Wolf Ketter
Email