Eye tracking as a tool for analysing human -AI image interactions

Abstract

In recent years, artificial intelligence (AI) has significantly advanced fields like computer vision, image description, and generation, proving particularly relevant in creative areas such as generative art. This research aimed to explore AI’s capabilities in creating and describing images compared to human perception. It included a comparative analysis of visual perception using eyetracking techniques in two settings: a VR art gallery created for the BITSCOPE project and a stationary ET study of individual images. The images, sourced from the BITSCOPE project’s CHIST-ERA IV collection, were initially described by an expert following specific instructions, which were then used by AI to generate corresponding images. The eyetracking study focused on key areas and gaze plot sequences, using a gaze plot similarity metric based on topology and path length, enabled by the size of the research group.