Evaluating Models of Fixation Selection
Author Information
Author(s): Wilming Niklas, Betz Torsten, Kietzmann Tim C., König Peter
Primary Institution: Institute of Cognitive Science, University of Osnabrück, Osnabrück, Germany
Hypothesis
How can we effectively evaluate models of fixation selection in eye-tracking studies?
Conclusion
The study concludes that the Area Under the Curve (AUC) is the most suitable measure for evaluating models of fixation selection, while also emphasizing the importance of considering both spatial bias and inter-subject consistency.
Supporting Evidence
- The study provides analytical proofs for the linearity of the ROC measure.
- It discusses the desirable properties for evaluation measures in fixation selection.
- The AUC is recommended as the best measure for evaluating models of fixation selection.
- The study highlights the importance of inter-subject consistency in model evaluation.
Takeaway
This study looks at how our eyes choose where to look when we see something, and it finds the best way to measure how good different models are at predicting those choices.
Methodology
The study reviews various measures for evaluating fixation selection models and proposes a framework for comparison based on AUC and KL-divergence.
Potential Biases
Potential biases may arise from the selection of subjects and the specific tasks they perform during eye-tracking.
Limitations
The study's findings are limited by the specific datasets used and the inherent variability in eye-tracking data.
Participant Demographics
48 subjects aged 19-28, with normal or corrected-to-normal vision.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website