Asymmetric impact of perception of attribute differences: random regret-based approach

Recently, regret-based choice models, which postulate that individuals systematically compare attribute values of pairs of alternatives in terms of the amount of regret that may be generated, and then choose the alternative that minimizes their anticipated regret, have attracted the attention of transportation researchers as an alternative to utility-maximizing models. Most studies have assumed that regret depends on attribute differences in objective, physical space. Perception is not taken into account. However, an overwhelming amount of empirical evidence suggests that the relationship between perceived and objective attributes is non-linear. Therefore, in the authors' previous work, the authors explored the predictive performance of random regret minimization models that incorporate the perception of attributes, based on the (generalized) Weber law. In this study, the authors expand their previous work by allowing asymmetric non-linear perception functions. The authors investigate the existence and nature of asymmetry for small and large attribute intensities. The proposed model specifications of the regret choice models are tested using two data sets. Estimation results and out-of-sample validation tests indicate that incorporating asymmetric perception functions of objective attribute differences significantly improves the predictive power of the considered regret-based choice models.