講演情報
[5O2-IS-5a-05]Modelling Attention and Mind-Wandering During Web BrowsingGoal Activation Dynamics and Memory-Driven Thought Transitions in ACT-R
〇Joy Karmoker1, Nethmee Rasanga Gunasekara1, Nilupul Heshan Randika Kodikara1, Umito Sorita1, Junya Morita1 (1. Shizuoka University)
regular
キーワード：
Cognitive Modeling、Mind Wandering Detection、Human Attention Dynamics、Web Browsing Behavior、Intelligent Learning Support
Maintaining attention during web browsing becomes difficult as motivation declines, often leading to mind-wandering. This study proposes a computational model implemented in ACT-R (Adaptive Control of Thought–Rational), a cognitive architecture that simulates human cognition through interacting memory systems and production rules. The model explains attention as cyclic goal competition driven by activation dynamics, memory retrieval, and noise. Web content is transformed into clustered semantic representations and stored in declarative memory, allowing alignment between task goals and browsing content. As goal activation decreases, off-task retrieval becomes more likely, producing mind-wandering episodes. Model behavior is evaluated using human data from a 30-minute web-based report-writing task. Page dwell time is compared with simulated quitting behavior. Results show that incorporating mind-wandering dynamics improves correspondence with human disengagement patterns.