Schedule of Events | Symposia

Sketchpad Series

Gaze reinstatement during naturalistic viewing and memory retrieval in children, adults and artificial intelligence models

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Iryna Schommartz^1,2 (schommartz@psych.uni-frankfurt.de), Bhavin Choksi³, Gemma Roig^3,4, Yee Lee Shing^1,2; ¹Department of Psychology, Goethe University Frankfurt, ²IDeA – Center for Individual Development and Adaptive Education, ³Computer Science Department, Goethe University Frankfurt, ⁴Center for Brains Minds and Machines, Massachusetts Institute of Technology

Differences in cognition and perception during image viewing influence processing and memory of scene elements. Scan paths during scene perception may also provide insights into pattern completion for partially incomplete images. However, the extent to which eye-gaze patterns predict subsequent memory and how these patterns differ between children, adults, and artificial intelligence (AI) models remains unclear. To investigate, we measured gaze fixations in children (aged 5–12) and young adults (aged 19–30) while they viewed 60 naturalistic images. Gaze fixations were then recorded during image reinstatement on a blank screen, cued by partially occluded images. Representational similarity analysis of fixation-based heat maps revealed that adults exhibited higher encoding-retrieval eye-gaze reinstatement than children, correlating with greater memory accuracy. This finding reflects the prolonged developmental trajectory of gaze reinstatement and its role in scan path consolidation. Using MultiMatch, a metric evaluating scan path similarity, we observed consistent differences between children’s and adults’ scan paths, highlighting developmental variations in scene perception. Additionally, we employed AI models to investigate their ability to predict scan paths of specific age groups. Providing models with initial human fixations or the first 10 seconds of gaze data significantly improved their performance in replicating mnemonic gaze reinstatement. Our findings have implications for cognitive neuroscience and the development of foveation-based AI models. They shed light on age-related differences in gaze behavior and offer insights for designing AI systems that emulate human-like visual exploration and memory processes.

Topic Area: PERCEPTION & ACTION: Vision