Integrating eye-tracking and first-person interviews to link individual differences in gaze and subjective decision structure

Humans have idiosyncratic and stable eye-scanning tendencies when exploring visual scenes. For instance, some people are more drawn to semantic objects while others are more drawn to emotional content. These individual variations are usually considered as a "noise ceiling" for fixation predictability. To break this ceiling, we need to better understand what causes these differences, and whether similar viewing tendencies are also reflected in similar cognitive profiles in observers. To this end, we tracked eye movements during an image-choice task in which subjects selected images varying in arousal. We then conducted first-person interviews (micro-phenomenological and cognitive analysis) to understand their decision process and identify individual cognitive strategies. We provide a proof-of-concept bridge between first-person elicitation and eye-tracking, demonstrating that structured experiential profiles align with objective oculomotor indices. Quantitative variables were derived from qualitative first-person interview verbatim, and then analyzed by Principal Component Analysis. We identified gaze parameters stable across trials and correlated them with the first three Principal Components. For instance, the gaze of participants who were more emotionally engaged was more driven by bottom-up visual saliency, and was more scattered across their preferred image. Our method uses the idiosyncrasy of gaze patterns as a structured signal reflecting personal decision process. We argue that observer-level factors derived from cognitive interviews can be integrated with image features to create individualized models of attention.