Elicitation of User Preferences by Cross Modality Matching

The development of autonomous multi-agent systems acting upon the interests of one or more users involves the elicitation of preferences of these users. In decision theory, models have been developed with which one is able to perform this elicitation. This paper enumerates some common anomolies to existing decision theoretic preference elicitation models. We propose an alternative elicitation model that has been successfully implemented and used for investigating positive time preference; this is the phenomenon that people prefer to get things sooner rather than later. We present new results concerning this phenomenon regarding decisions in gain and loss situations over time. The contributions of this paper are (1) the presentation of our novel preference elicitation model, and (2) new results on the phenomenon of gain-loss asymmetry in intertemporal choice.