Hauptinhalt

FB Kolloquium 2023: Prof. Dr. Jella Pfeiffer

„Consumer Behavior Analysis and Consumer Decision Sup-port Using Eye-Tracking Data”

Veranstaltungsdaten

Prof. Dr. Jella Pfeiffer, Universität Gießen
„Consumer Behavior Analysis and Consumer Decision Support Using Eye-Tracking Data”

The retail sector is undergoing a transformation towards virtual commerce and the process lately gained momentum by the rise of the metaverse idea. If such an interconnected platform of 3D world takes shape, future customers can experience products and services in a plethora of augmented and virtual environments. If this development realizes, soon many consumers’ eye-tracking data will be available because most of the devices needed for accessing augmented and virtual reality include eye trackers. In this talk, we present some of our studies which combine customer behavior research, machine learning and the application of virtual reality. The first presented study focusses on two frequently investigated shopping motives: goal-directed and exploratory search. To train and evaluate a prediction model, we conducted two eye-tracking experiments in front of supermarket shelves. Our empirical results show that shallow machine learning algorithms allow correct classification of search motives with high accuracy both in physical and virtual reality just based on the consumers’ eye-tracking data. Our findings also imply that eye movements allow to identify shopping motives relatively early in the search process. The second presented study shows how gaze information in virtual reality applications might soon serve as features for recommender systems which learn user preferences on-the-fly. Drawing on the eye-mind hypothesis, we imagine a recommender system that predicts customer health-consciousness. We propose to treat gaze information as time series and use a deep time series classifier for inference. Classification results indicate superior performance compared to a shallow baseline. In a third study, we investigate good timing to approach customers with digital sales agent or with support systems and propose a machine learning model to automate this process. Our findings suggest that cognitive load can serve as predictor to determine good invocation timing for algorithmic user assistance.