Socially rational agent-based model for group decision-making

What is it about?

This study employs a computationally effective Bayesian learning technique, along with social value orientation theory, to design socially rational intelligent agents who work on behalf of real actors in an urban land use planning casework.

Featured Image

Why is it important?

Given that the aim of decision support systems is to help users resolve decision problems, it is especially important to model user preferences. If no information is available at the beginning of the interaction, preference elicitation methods must attempt to obtain as much information about users’ preferences as possible, so that the systems can help users move toward their goals (Chen and Pu 2004). Although there is a vast array of work that has tried to address different parts of the user preference elicitation problem (for a review, see Chen and Pu (2004) and Braziunas and Boutilier (2009)), to the best of our knowledge, no studies have yet proposed the use of intelligent agents that elicit users’ social preferences and utilize them to work on behalf of the users. Therefore, this paper describes the incorporation of a specific Bayesian learning and social value orientation (SVO) theory in order to retrieve actors’ social preferences and simulate the negotiation process in the context of a multi-actor ULUP. This work represents a follow-up to a previous study conducted by Ghavami et al. (2016) which investigated the feasibility of using socially rational agents to simulate a ULUP. The current study proposes the use of Software Intelligent Agents (SIAs) equipped with a Bayesian learning mechanism to learn about the SVO of associated human actors. It also employs an approximation method based on the sequential processing and updating of preference parameters as well as the systematic sampling of parameter spaces in order to reduce the learning process’s computation time.