Eye-tracking-based hidden Markov modeling for revealing within-item cognitive strategy switching.

Identifying cognitive strategies in problem-solving helps researchers understand advanced cognitive processes and their applicable contexts. Current methods typically identify strategies for each item of Raven's Advanced Progressive Matrices, capturing only between-item cognitive strategy switching (CSS). Although within-item CSS is recognized, methods to dynamically identify and reveal it are lacking. This study introduces the concept of an eye movement snippet, a basic unit for studying within-item CSS, along with a new eye-tracking process measure that quantifies the sequence length of alternatives viewed in a snippet. Combined with hidden Markov modeling, we propose a new method for dynamically identifying within-item cognitive strategies and revealing their switching. Using eye-tracking data from a matrix reasoning test, we demonstrate the value of the proposed method through a series of analyses. The results indicate that during problem-solving: (1) participants predominantly used two strategies-constructive matching and response elimination; (2) there is a high probability of switching from constructive matching to response elimination, but not vice versa; (3) more difficult items lead to more frequent strategy switching; (4) frequent strategy switching decreases time spent in the matrix area and on problem-solving planning; (5) frequent strategy switching correlates with incorrect answers for some items; and (6) frequent strategy switching increases total response time. Additionally, within-item CSS showed three distinct patterns as the test progressed, with significant differences in participants' intelligence levels and total test time among the patterns. Overall, the proposed method effectively identifies within-item cognitive strategies and their switching in matrix reasoning tasks.