Performing Department
(N/A)
Non Technical Summary
Healthy food choices are central to nutritional security, but these choices require individuals to find a complex balance between interests in the present and future. Therefore, policies to encourage healthier eating must consider food characteristics to which people give greater attention when making these intertemporal decisions and the psychological mechanisms through which those characteristics influence the decisions. In this project, we propose to study emotion as one such psychological mechanism.We propose to utilize cutting-edge machine vision and artificial intelligence technologies to evaluate the role of attention and emotion in food choice. These technologies use cameras to track subjects through a physical space, measure attention through eye-tracking (what information is currently being processed), and measure emotion through facial micro-expressions (how that information affects a subject's emotional state).We will investigate the role emotions play in food choice decisions. We are also interested in how the relationship between emotion, attention, and food choice is affected by whether the food-choice environment is digital (e.g. online shopping) or physical (e.g.a grocery store). We will use both acomputer laboratory and a more naturalistic, simulated shopping experience to test these questions.
Animal Health Component
33%
Research Effort Categories
Basic
34%
Applied
33%
Developmental
33%
Goals / Objectives
We propose to utilize cutting-edge machine vision and artificial intelligence technologies to evaluate the role of attention and emotion in food choice. These technologies use cameras to track subjects through a physical space, measure attention through eye-tracking, and measure emotion through facial micro-expressions. We propose three key research questions: 1) How do emotions mediate food choice decisions? 2) How well do emotion and attention predict food choices in a natural environment? And 3) How does emotion interact with context to influence food choice?Our proposed design evaluates these questions in two phases--first in a computer laboratory and subsequently in a more naturalistic, simulated shopping experience. We aim to identify attention and emotion variables that mediate the relationship between nutritional attributes and food choice. We will also observe novel details about food choice behaviors by combining advanced technology with our controlled, yet naturalistic, shopping experience. We further propose to explore how measures of emotion predict heterogeneity in responses to randomized interventions. Finally, we will evaluate the role of emotion in determining how behavior is influenced by the food-choice environment.
Project Methods
Computer Science Methods: Our project will take existing computer vision algorithms thatdetectfacial microexpressions andtrackeyegaze and tailor them to a food-choice environment.Experimental Methods to ElicitFood Choice: Subjects will participate in food-choice experiments where nutrition interventions (such as information, context, or price)arerandomly varied. This randomization methodology is straightforward: a predetermined fraction of subjects will receive each intervention and these subgroups will be assigned randomly. These food-choice experiments will take place digitally in front of the computer and in-person in the aisles of a simulated grocery store.Statistical Methods to Analyze Data:We will use mediation analysis to determine the role that emotion plays in determining how food choices respond to food characteristics. Through this process, we will determine the "total effect" that food characteristics have on food choices and decompose that into the "direct effect" and the "indirect effect" that is mediated through emotion. This will follow the mediation analysis in Baron and Kenny (1986).We will perform this analysis for a suite offood characteristics and six emotions. Specifically, for each food characteristic, we will identify the emotions that mediate its effect on food choice. In this way, we can separately categorize food characteristics based on the emotions most closely associated with them. For each emotion, we can establish its impact on food choice through the regression coefficient. We can transform this slope into an elasticity by re-scaling using mean levels of each emotion and the mean food choices.We will estimatethe causal impacts of our experimentally-random interventions through astraightforward comparison of means. We willestimate the role emotion plays in mediating these treatment effects by includinginteraction terms to determine whether the effectiveness of interventions is predicted by the emotions triggered.