Restricted Intervals in Preference Dynamics: Theory and Application in an Agent-Based System

Agent-based models have long been used to generate states of variables at individual and aggregate levels. More recently, robotic and computer agents are being given forms in which internal dispositions such as cognitive reasoning, emotions, and preferences are represented. We note limitations in most common representations of preferences in these applications and propose conditions for dynamic forms of the variable. A consequence of the dynamics we propose is that preferences exhibit continuous variation in restricted intervals while maintaining characteristic differences between entities and over time. After giving a form to preferences in a situated application to a consumption system, analytical and computational methods are used to establish that dynamics in the system have properties we describe. We then derive dependencies of the restricted intervals in which preferences typically vary. Estimation of these intervals is exemplified in data on the U.S. economy. The system is then embodied in a network form in which cycle and randomness from environments and their interactions are represented. A demonstration of stochastic resonance in the networked system supports the imputed dynamics. The network model is also used to indicate network dependencies of the bounds of the interval in which preferences vary. Finally, the relevance of the dynamics that are proposed in preferences to other internal dispositions that include emotion in agent-based models is discussed.