The Dynamics of Attribute Weighting in Multiattribute Choice

How much flavour do we give up in pursuit of a healthier choice, and how many extra minutes of commute do we sacrifice in fostering a more sustainable future? Daily decisions involve navigating multiattribute tradeoffs by assigning weights to different attributes. Understanding the mechanisms underlying attribute weighting is key to explaining and predicting choice behaviour. However, these mechanisms are poorly understood. Past research has examined how people adjust attribute weights in response to shifts in goals or in the context. Beyond induced shifts, it is unknown whether attribute weights maintain stability in stationary contexts, from one decision to the next. My central hypothesis is that, akin to other representations in the brain (e.g., memories), attribute weights are in constant flux, being controlled by inherently noisy and dynamical processes. This project aims to unravel these processes.Attribute weights are inherently subjective and not directly observable. Reading them out from single choices is impossible, as choices are perturbed by noise and influenced by normatively irrelevant factors. To overcome this limitation, I will use multiattribute choice tasks with magnetoencephalography and, focusing on decision-relevant neural signals, I will separate the influence of attribute weights from the distorting influence of irrelevant factors. This will enable me to read-out attribute weights at each decision, and to subsequently chart their temporal dynamics across successive multiattribute decisions.Understanding the dynamics of attribute weighting will offer mechanistic insights on how preferences change, informing century-old debates on the nature of preferences. These insights will be valuable to clinicians seeking to understand pathological preference variability or persistence (e.g., in addiction); and to policy makers probing whether specific consumer tradeoffs (e.g., between monetary and environmental attributes) are stable or changeable.