An emotional-based methodology to detect preferences in a decision-making process applied to a virtual service robot

Abstract

In a multi-objective problem, no single solution optimally satisfies all objectives. Thus, the challenge is to find a balance between conflicting objectives. The decision-making necessarily requires human intervention. The person responsible for selecting the most appropriate solution among all the trade-off solutions is the decision maker (DM). The DM seeks to approach only the solutions that best suit her/his preferences. Since there is plenty of specialized literature showing that emotions play a critical role in decision-making, we aim to incorporate them into the decision-making process. To elicit emotions that can be quantifiable, we propose the Emotional Assessment Method. The method presents a simulation of the objectives to be optimized that represents the consequences of each decision. Using this methodology, the decision maker assesses the emotions evoked by each presented simulation to guide his/her search for solutions that satisfy his/her preferences. As a case study, we aim to identify subjects’ preferences towards robot behaviors. Seventy-two subjects with varying levels of familiarity with robots (divided into two datasets) participated in the experiments. We concluded that the method elicits subjects’ emotions while observing the consequences of the robot’s performance. Also, we found out that it is possible to identify subjects’ preferences based on both the context and the emotions to select the robot’s behavior.