Related Experiment Video
Updated: Jun 14, 2025
07:09
Gaze in Action: Head-mounted Eye Tracking of Children's Dynamic Visual Attention During Naturalistic Behavior
Published on: November 14, 2018
10.6K
Gaze-Based Detection of Thoughts across Naturalistic Tasks Using a PSO-Optimized Random Forest Algorithm
Tarannum Rahnuma1,2, Sairamya Nanjappan Jothiraj1,2, Vishal Kuvar3
1Department of Psychology, University of Calgary, Calgary, AB T2N 1N4, Canada.
Bioengineering (Basel, Switzerland)
|August 29, 2024
Summary
Researchers used eye tracking and machine learning to classify nine dimensions of ongoing thoughts during naturalistic computer tasks. This approach shows promise for real-time prediction of human cognition in everyday settings.
Area of Science:
- Cognitive Science
- Neuroscience
- Human-Computer Interaction
Background:
- Ongoing thoughts are central to human experience and influence mood, well-being, and productivity.
- Previous research linked eye movements to specific thought dimensions (e.g., task-relatedness) in experimental settings.
- Limited understanding exists on classifying diverse thought dimensions via oculomotor activity during naturalistic activities.
Purpose of the Study:
- To determine if various thought dimensions can be classified using oculomotor activity during naturalistic tasks.
- To assess the efficacy of machine learning models in predicting thought dimensions from eye movement data.
- To explore the potential for real-time prediction of ongoing thoughts in ecologically valid contexts.
Main Methods:
- Utilized thought sampling, eye tracking, and machine learning across seven participants over multi-day, self-selected computer tasks.
- Collected 1715 thought probes from 63 hours of recordings.
- Employed a random forest (RF) classifier with particle swarm optimization (PSO) for feature selection, using fixation and saccade data.
Main Results:
- Achieved above-chance classification performance for all nine assessed thought dimensions (task-relatedness, freely moving, stickiness, goal-directedness, internal-external orientation, self-orientation, others orientation, visual modality, and auditory modality).
- Mean Matthews correlation coefficient (MCC) values ranged from 0.25 to 0.54.
- PSO-based feature selection improved RF classifier performance compared to using all features.
Conclusions:
- Machine learning combined with eye movement measures can classify multiple dimensions of naturalistic ongoing thoughts.
- This approach demonstrates potential for real-time prediction of cognitive states in real-world settings.
- Oculomotor activity provides valuable insights into the complex stream of human consciousness.
