Adaptive Decision Making Under Salience Effects

Sammanfattning

When people (repeatedly) choose among items characterized by multiple attributes ( price, quality, brand etc..) they often over-or under-weight certain attributes depending on which features stand out in the choice set. Salience theory explains how these distortions arise in a static setting. I extend the existing salience theoretic model for deterministic choice to a setting in which the decision maker chooses from the same choice set several times adapting his reference and preference. Furthermore, I distinguish the utility they anticipate when deciding from the utility they actually experience, letting both the reference point and pre-distortion weights adapt over time, based on the difference between these two notions. The main analytical results regard the convergence of reference points to a weighted average of items that a rational agent could choose. I specify conditions which allow this result and complementary results. I further analyse the proposed dynamic mechanism through simulations which show that all agents benefit from salience in this setting. Adaptation regarding reference points and preferences are both beneficial, especially when combined. I compare standard salience theory with salient regret theory (a decision rule with attractive properties) and find that the prior yields higher experienced utility than the latter in simulations. Lastly, I make a stronger observation about the convergence of reference points based on simulation results.