Computational modeling of health decision-making
No Thumbnail Available
Authors
Meeting name
Sponsors
Date
Journal Title
Format
Subject
Abstract
Value-based decision making consists of weighting stimulus attributes to compute stimulus value and guides choice such that the stimulus with highest value is chosen. While successful in maximizing value in the short-term, this does not always lead to long-term advantageous choices. Incorporating self-control, the ability to choose long-term rewards over short-term rewards, into the decision process guides choices to be congruent with one’s long-term rewards. Therefore, individuals must utilize self-controlled value-based choice which guides choice to maximize value in alignment with long-term rewards. Self-controlled value-based choice successfully captures the decision dynamics in food decision-making. Little has been done to see if self-controlled value-based choice captures decision dynamics in activity decision-making. Considering the role of food and activity choice in energy balance (calories in, calories out) it is possible they share a similar decision process. The present experiment examined the cognitive and physiological decision dynamics of self-controlled value-based food and activity choice using computational methods. Individuals completed a series of decision-making tasks for food and activities while electroencephalogram (EEG) and electrocardiogram (ECG) were recorded. EEG data were then input into a convolutional neural network (CNN) to classify self-control success/failure choices and classify individuals as high/low self-controllers. Results show a similar valuation process for food and activity stimuli by independently weighting stimulus attributes to compute value. To promote self-controlled value-based choices, the valuation process is altered by enhancing the weight of the long-term reward attribute (health) and diminishing the weight of the short-term reward attribute (taste/enjoyment). The CNN classified self-control success and failure choices and high/low self-controllers above chance level. The EEG electrodes that led to the highest classification accuracy were located over frontal, parietal, and occipital regions known to be involved in self-controlled value-based choice. Feature visualization revealed the networks place high importance on theory-aligned brain responses for classifying food data, but theory-unaligned brain response for classifying activity data. Overall, the experiment successfully modeled food and activity decision-making demonstrating both domains share a similar decision process and demonstrated the feasibility of applying advanced computational methods to EEG data to elucidate the physiological processes underlying the cognitions involved in the decision process.
Table of Contents
Introduction -- Review of literature -- Methods -- Results -- Discussion
DOI
PubMed ID
Degree
Ph.D. (Doctor of Philosophy)