BKT students completed posttest problems significantly faster than Control students, with DRL showing marginal advantages in time and solution optimality¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? BKT students completed posttest problems significantly faster than Control students (10.8 vs 11.8 minutes, p= 0.02, A=.58). → Tithi 2026 q3 i? DRL students showed marginally lower posttest completion time and marginally fewer steps (more optimal solutions) than Control (both p= 0.06). → Tithi 2026 (2)

Evidence¶

Tithi 2026¶

Tithi, S.D., Alam, N., Yasir, T., Shi, Y., Tian, X., Chi, M., Barnes, T. (2026). Adaptive Scaffolding for Cognitive Engagement in an Intelligent Tutoring System. arXiv. https://arxiv.org/abs/2602.07308 doi:10.48550/arxiv.2602.07308

q3 · i? · causal · r2

Posttest problem completion time comparison across conditions in the N=113 study; BKT mean 10.8 minutes versus Control 11.8 minutes with A=.58. DRL's time difference was marginal (p= 0.06, A=.56).

"students inBKTcompleted posttest problems signifi- cantly faster than students inControl(Control(mean) = 11.8 minutes per prob- lem,BKT(mean) = 10.8 minutes,p= 0.02,A=.58, 95% CI [.53, .63])"

Tithi 2026 (2)¶

Tithi, S.D., Alam, N., Yasir, T., Shi, Y., Tian, X., Chi, M., Barnes, T. (2026). Adaptive Scaffolding for Cognitive Engagement in an Intelligent Tutoring System. arXiv. https://arxiv.org/abs/2602.07308. doi:10.48550/arxiv.2602.07308

q3 · i? · causal · r2

Posttest solution optimality (steps) comparison; shorter solutions with fewer steps are more optimal in this domain. The DRL advantage over Control was marginal (p= 0.06) rather than significant.

"TheDRLgroup had marginally more optimal solutions with fewer steps than theControlgroup in posttest problems (Control(mean) = 8.5 steps,DRL(mean) = 7.7 steps,p= 0.06,A=.56, 95% CI [.51, .61])"

Discussion¶

Related Claims¶

• Adaptive scaffolding policies (BKT and DRL) significantly improve posttest performance over a non-adaptive control in a logic ITS — related
• Adaptive-condition students complete the posttest in significantly less time than Control students — a broader claim this one bears on
• Students receiving Adaptive proactive hints based on HelpNeed predictions achieve significantly higher posttest optimality than Control students — related
• The DRL policy produced a significantly different scaffolding distribution than Control and BKT, favoring Guided examples (60%) and avoiding Buggy examples (4%) — related
• No significant differences in posttest rule accuracy across scaffolding conditions — related
• Both hint conditions took more time-on-task than control, but ChatGPT and human tutor conditions did not differ in session time — related
• Retention and delivered instructional minutes did not differ significantly between experimental and control conditions, and completers received well under planned dosage — related