Eye-Based Recognition of User Traits and States-A Systematic State-of-the-Art Review

Area of Science:

• Human-Computer Interaction
• Machine Learning
• Cognitive Science

Background:

• Eye-tracking technology offers detailed insights into user visual behavior and interests.
• Machine learning advancements enable the recognition of user traits and states from eye-tracking data.
• A comprehensive review of eye-based recognition methods is currently lacking.

Purpose of the Study:

• To systematically review and synthesize machine learning-based approaches for recognizing user traits and states using eye-tracking data.
• To identify state-of-the-art methods, research gaps, and future directions in the field.
• To provide a conceptual framework for understanding eye-based user recognition.

Main Methods:

• Systematic literature review following PRISMA 2020 guidelines.
• Searches conducted in ACM Digital Library and IEEE Xplore.
• Inclusion of studies using eye-tracking data with machine learning or deep learning for user recognition.
• Risk of bias assessment using standard methodological criteria.
• Data synthesis incorporating a conceptual framework (task, context, technology, data processing, recognition targets).

Main Results:

• 90 studies were included, covering diverse tasks (visual, driving, learning) and contexts (screen, simulator, real-world).
• Recognition targets included cognitive/affective states (emotions, workload) and user traits (personality, memory).
• Various machine learning techniques (SVMs, Random Forests, deep learning) were employed.
• Key gaps identified include the need for best practices, larger datasets, and varied tasks/contexts.

Conclusions:

• Future research should enhance ecological validity and explore multi-modal approaches for robust user modeling.
• Development of gaze-adaptive systems is a promising future direction.
• Standardization of methods and data collection is crucial for advancing eye-based user recognition.