Gaze bias: Selective encoding and liking effects

Abstract

People look longer at things that they choose than things they do not choose. How much of this tendency—the gaze bias effect—is due to a liking effect compared to the information encoding aspect of the decision-making process? Do these processes compete under certain conditions? We monitored eye movements during a visual decision-making task with four decision prompts: Like, dislike, older, and newer. The gaze bias effect was present during the first dwell in all conditions except the dislike condition, when the preference to look at the liked item and the goal to identify the disliked item compete. Colour content (whether a photograph was colour or black-and-white), not decision type, influenced the gaze bias effect in the older/newer decisions because colour is a relevant feature for such decisions. These interactions appear early in the eye movement record, indicating that gaze bias is influenced during information encoding.

Keywords:

• Decision making
• Eye movements
• Gaze bias
• Liking
• Selective encoding

Acknowledgements

This research was supported by grants HD26765 and HD51030 from the National Institute of Health. We thank Julian Parris, Timothy Slattery, and Bernhard Angele for help with analysis and preparation of the data, Simon Liversedge, Eyal Reingold, Mackenzie Glaholt, and two anonymous reviewers for helpful feedback on a previous draft of the paper, and a number of colleagues at UCSD for their feedback at informal presentations of the data.

Notes

¹The reason why there is not an equal number of pairs which have the same colour content and different colour content is because we originally planned for the same colour content pairs to be filler items to ensure that the task was not too easy, but later realized that the comparison would be interesting between same colour content pairs and different colour content pairs.

²We ran separate ANOVAs on the like/dislike and older/newer data instead of using liking vs. nonliking decisions as another factor because the distinction between like and dislike is qualitatively different than the distinction between older and newer. Therefore, our design is not really a 2(decision type: Liking vs. nonlike decision)×2(attribute: Positive vs. negative) design×2(gaze bias effect: Chosen vs. not chosen), but rather a 2(decision type)×2(gaze bias effect) design in which one comparison (liking vs. nonliking decisions) has a nested variable within it (like vs. dislike and older vs. newer).

³Despite the larger numerical difference between the means for the like and dislike condition compared to the difference between the chosen and not-chosen item, there was much greater variance in the dwell durations between the like and dislike condition, leading to greater error variance in the ANOVA for decision type. Therefore, the main effect of decision type was not significant while the main effect of whether the item was chosen was significant.

⁴We used the corrected F-values because there was a slight violation of sphericity.