Search papers, labs, and topics across Lattice.
This study analyzes clustering of intelligent tutoring system logs from 5,000 learners on EdNet-KT3 to explore the relationship between study strategies and engagement rather than mastery. By identifying eight distinct study-strategy clusters, the researchers found that early engagement patterns significantly predicted later persistence and session completion, but did not correlate with unassisted accuracy in late attempts. The results challenge the assumption that learner types derived from behavior logs can effectively predict knowledge acquisition, highlighting a disconnect between engagement and mastery outcomes.
Early engagement strategies predict persistence in learning sessions, but they don't correlate with actual knowledge gains. WHY_IT MATTERS: This insight could shift how educational technologies leverage behavioral data to enhance learner outcomes, focusing more on engagement than mastery.
Learning analytics often treats unsupervised clusters of intelligent tutoring system (ITS) logs as learner types that should predict learning. We test that assumption on EdNet-KT3. Clustering study-strategy features (resource use, revision, video, problem practice) for 5{,}000 active learners yields a silhouette-selected parent cut ($k=5$) with 4 contrast poles (reading-focused, video-heavy, revision-heavy, and problem-first) plus a large near-mean residual ($\sim$64.9\%). Reclustering that residual adds four finer styles, giving a bootstrap-stable hierarchy of 8 named strategies. We split each learner's timeline by respond count so clusters use only the early half and outcomes only the late half. Early clusters predict later engagement (continuing to practice and finishing late sessions, especially persistence, $畏^{2}\approx 0.106$; completion $畏^{2}\approx 0.021$) but not later unassisted accuracy (correctness on late first-attempts without help; $p_{\mathrm{adj}}\approx 0.093$). Volume rises with some styles, yet volume-only clustering barely matches strategy labels (ARI$=0.064$). A knowledge-tracing model (SAKT) on the seven TOEIC exam sections predicts next correctness only modestly better than a baseline that knows only how hard each section usually is (AUC lift $+0.051$; CI $[+0.045,+0.058]$), and that mastery signal is nearly independent of behavior styles (ARI$=0.007$). Behavioral clustering here describes study styles and engagement, not knowledge gains.
