Attribute commensurability and context effects in preferential choice

Abstract

Context effects in multi-alternative, multi-attribute choice are widely documented, but often elusive. We show that this elusiveness can arise in part from the way that choices are presented. To illustrate this, we use a modeling framework to predict how changes to the format of attribute values, specifically the commensurability of attribute values, influences attention allocation and consequently context effects. Guided by this framework, we show in two online choice experiments (total N = 954 adults) that manipulating the commensurability of attributes leads to different patterns of context effects. Robust attraction and compromise effects are found when attributes are incommensurable (e.g., CPU speed in GHz and RAM memory in GB, or quality ratings on different scales), and mostly null effects occur when attributes are commensurable (e.g., quality ratings on the same scale). Our findings show how the format of choice information can substantially alter the integration of that information and resulting choice patterns.