Eye-Tracking Study Reveals Cognitive Load of AI Disclosure in News
A new study on arXiv (2605.14999) explores how AI disclosures in journalism affect reader engagement and mental effort. The researchers conducted an experiment with three main variables: the level of disclosure (none, one line, detailed), the type of news (politics or lifestyle), and the role of AI (editing or partial content creation). By using NASA-TLX and eye-tracking methods, they found that one-line disclosures significantly increased how long readers focused on the text and how often they moved their eyes, especially in AI-edited pieces, indicating a heavier cognitive load. In contrast, detailed disclosures didn’t seem to add any extra burden. The study relates this to Information-Gap Theory, suggesting that brief disclosures can prompt more critical evaluation from readers.
Key facts
- Study published on arXiv with ID 2605.14999
- Mixed factorial design: 3x2x2
- Independent variables: disclosure detail (none, one-line, detailed), news type (politics, lifestyle), AI role (editing, partial content generation)
- Dependent measures: NASA-TLX and eye-tracking metrics
- One-line disclosures led to higher fixation durations and saccade counts
- Effect was stronger for AI-edited content
- Detailed disclosures did not impose additional cognitive load
- Theoretical framework: Information-Gap Theory
Entities
Institutions
- arXiv