Human classifier: Observers can deduce task solely from eye movements.

Area of Science:

• Cognitive Science
• Human-Computer Interaction
• Visual Perception

Background:

• Computer algorithms effectively classify tasks using eye movement data.
• Human ability to classify tasks from eye movements remains largely unexplored.

Purpose of the Study:

• To investigate whether humans can accurately classify cognitive tasks based solely on observed eye movements.
• To determine the influence of different eye movement representations (fixations, scanpaths, videos) and scene context on human classification accuracy.

Main Methods:

• Two experiments were conducted where human participants classified tasks (Search, Memory, Rating) based on presented eye movement data.
• Eye movement data were visualized as fixations (circles), scanpaths (arrows), or videos (moving dot), with and without the original visual scene.

Main Results:

• Participants successfully classified the Search task, with highest accuracy when eye movements were shown without the scene, particularly fixations.
• The Memory task was also classified above chance, showing strongest performance with video representations of eye movements presented within the original scene.
• Analysis revealed specific eye movement properties utilized by successful human classifiers.

Conclusions:

• Humans can extract task-relevant information from eye movement characteristics under specific conditions.
• The effectiveness of human eye movement classification depends on the task, the representation of eye movements, and the presence or absence of contextual scene information.
• Findings provide insights into human visual attention and inform the development of more effective computer-based eye movement analysis systems.