Context Effects from Stimulus Generalization: A Similarity-Weighted Random Utility Model

Context effects arise when adding or removing options changes preferences among the remaining options. This paper explains these effects by bringing stimulus generalization from animal learning into random utility theory. The Similarity-Weighted Utility (SWU) model combines two processes that affect the systematic utility of the options: value spreads between similar options, whereas very similar options also compete with each other, with choice following the usual logit rule. SWU produces attraction, compromise, and similarity, and the same setting also accounts for effects of decoy position, an increase in preference for a target when a superior unavailable option is presented, and background contrast when earlier options remain represented in memory. Further simulations examine changes in discrimination and violations of absolute regularity. The model also makes new predictions. Improving every option by the same amount can reverse a compromise preference while leaving similarities and all binary choices unchanged. An unavailable negative option can also shift the peak of preference away from itself when equally valued options are tested one at a time. These results show how value generalization and competition can explain context effects within a random utility framework, connecting animal and cognitive psychology with economic choice modeling.