The heterogeneous heuristic modeling framework for inferring decision processes

Purpose – Increasing evidence suggests that choice behaviour in real world may be guided by principles of bounded rationality as opposed to typically assumed fully rational behaviour, based on the principle of utility- maximization. Under such circumstances, conventional rational choice models cannot capture the decision processes. The purpose of the chapter is to propose a modeling framework that can capture both decision outcome and decision process.Methodology – The modeling framework incorporates a discrete cognitive representation structure and implies several decision heuristics, such as conjunctive, disjunctive and lexicographic rules. This allows modeling unobserved decision heterogeneity involved in a single decision, for example, in the form of a latent-class specification, taking into account mental effort, risk perception and expected outcome as explanatory factors.Findings – Two models based on this framework are applied to decision problems underlying pedestrian shopping behaviour and compared with conventional multinomial logit models. The results show that the proposed models may not be superior to logit models in terms of model selection criteria due to the extra complexity in selecting heuristics, but suggest more interesting insights to the underlying decision mechanisms.Research implications – Understanding decision processes additional to outcomes is a promising research direction. A more developed model should take into account more contextual and socio-demographic factors in the heuristic selection part. The assumptions of information processing must be subject to empirical tests to validate the model.Originality – The proposed modeling framework bridges the long-existing contradicting approaches in the field of decision modeling, namely the rational approach and the bounded rational approach, by proving that noncompensatory decision heuristics can be inferred from compensatory model formulations with discretized information representations and decision criteria assumed. It also incorporates a heuristic choice part into the decision processes in the form of latent-class specifications and shows the viability of the new modeling framework.