Mapping smartphone-based eye-tracking behavior across Japanese individuals on the pareidolia test

Abstract

AbstractPareidolias are illusionary phenomena wherein ambiguous forms appear meaningful. In clinical research, pareidolias have been studied using paper or desktop test formats to deconstruct visuo-perceptual mechanisms. Translating this work on to an accessible, scalable setup such as smartphones is currently unknown. Here, we designed a smartphone-based pareidolia test to study visual processes affecting gaze behavior of cognitively healthy individuals using a standard, native front-facing camera. We optimized our system using machine learning and explored the challenges involved in user behavior, demographic specificity, and test functionality. We performed our experiments on 52 healthy Japanese adults, aged between 50 to 80 years who underwent MMSE and the smartphone test for pareidolias. Gaze movements on the 15-min, user-centric evaluation was calibrated to every individual. Results showed test responses with minimal differences with respect to age, sex, and completion time. Personalized calibrations improved the model’s prediction performance and quantification of gaze tracking metrics aligned with that of commercial grade eye-trackers. Our findings demonstrate the applicability and scalability of pareidolia testing on smartphone platforms.