Files
Abstract
Most of the decision aid agents are designed for specific domain problems. These agents are based on different approaches based on the specific needs of the decision problem. Our research is on developing a decision-aid system as an agent called 'Rae' which will attempt to represent any decision problem as a POMDP in an intuitive way. We model the approach of solving any decision problem through a decision-making agent designed to help users elicit their decision problems. We present approaches for elicitation of user preferences into reward function of POMDP based on Analytical hierarchical process, symmetry of actions, elicitation of transitions, elicitation of observations, policy representation and designing an interactive text/voice based agent for POMDP. We present hypotheses of a decision making agent and evaluate Rae by conducting a human usability study.