Predicting food choices based on eye-tracking data: Comparisons between real-life and virtual tasks

As eye-trackers become increasingly important in studies on consumer food choice, it is crucial to test the ecological validity of virtual eye-tracking tests. The present study aims to cross-examine eye-tracking data obtained from real-life versus virtual food choice tasks. Sixty-two healthy females participated in this study by attending two sessions, with virtual and real-life settings, respectively. Both sessions were constructed identically - with participants required to view eight common snack food items with different arrangements, while wearing mobile eye-trackers. To complete each task, the participants were asked to select three types of food for consumption. Analyses of summed dwell time (i.e., total visit duration - the summed latency of gaze visit on an 'area of interest' from entry to exit) were performed to assess food attention biases across test conditions, and between the selected and unselected food items. While the results revealed only minor differences in visual preferences between real-life and virtual settings, data from these two settings showed differential relationships to food choices. Eye-tracking data obtained in the virtual setting supported the notion that the selected food items were looked at longer (p < 0.05). However, the dwell time is shown inadequate to fully capture the more complex cognitive processes underpinning real-life food choices, with non-significant differences being reflected in dwell time for selected versus unselect foods. Overall, our study demonstrates inconsistent outcomes of eye-tracking food research in virtual versus real-life settings, highlighting the importance of accounting for environmental variation when interpreting eye-tracking data for food cues.

Automated summary

This study compared eye-tracking data obtained in real-life and virtual settings, finding that the selected food items in the virtual setting were looked at for a longer duration, indicating a differential relationship between data from these two settings. However, dwell time is insufficient to fully capture the complex cognitive processes underlying real-life food choices, with non-significant differences in dwell time for selected versus unselected foods.