ISR Colloquia Series: Dana Nau, "State-Dependent Risk Preferences in Social-Modeling Games"
Wednesday, May 5, 2010
1146 A.V. Williams Building
301 405 6615
Evolution of State-Dependent Risk Preferences in Social-Modeling Games
Professor Dana Nau
Institute for Systems Research and Department of Computer Science
University of Maryland
There is much empirical evidence that human decision-making is state-dependent: humans are sometimes risk-averse and sometimes risk-seeking, depending on their current situation. This differs from the decision-theoretic notion of expected-value maximization, in which a decision-maker prefers the alternative with the highest expected value, regardless of its risk.
An important open question is what evolutionary pressures might have caused state-dependent risk preferences to become prevalent, and I'll describe some work that my students and I have done to investigate this question. Our analysis and simulations show that in evolutionary game environments where an agent must make a sequence of choices among alternatives that have differing amounts of risk, state-dependent risk preferences can have an evolutionary advantage over expected-value maximization.