Modeling Agent's Preferences Based On Prospect Theory

It is well known that human preferences in decisions under risk do not always complies with expected utility theory (EUT). In fact, there are several effects that are inconsistent with basic tenets of EUT. Alternative theories have been proposed and perhaps the most well studied is Prospect Theory (PT). Recent work showed experimental results that support the idea that financial professionals may behave according to PT and violate EUT. Meanwhile, some argue that economy needs agent-based modeling, because it may be a better way to help guide financial policies than mathematical models. If financial professional behave according to PT in markets, then agent-based modeling needs PT based agents. Our idea is creating trading agents based on PT to simulate a market. However, the creation of an artificial agent based on PT as originally proposed is very hard and limited to two outcome prospects. We propose an agent model based on an extension of PT called Smooth Prospect Theory (SPT). We used this model to create agents to populate an artificial market with SPT and EUT agents. It was used to predict real market behavior for short periods. SPT agents provided more accurate predictions in crisis periods than EUT agents.