The association between RT propensity and learning rate is moderated by prior proficiency: slower propensities predict greater learning efficiency for higher-proficiency students but are weakly negative for lower-proficiency students¶

Claim · All claims
Evidence · 2 studies · 2 associational r2 · q2
Currency · no dated study recorded · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

q2 i? There is a significant RT propensity × prior proficiency interaction predicting learning rate (β = 0.11, p = .004), with longer RT propensities associated with steeper opportunity slopes among higher-proficiency students and weakly negative among lower-proficiency students. → interaction-1 q2 i? The interaction remains significant in a sensitivity analysis excluding potentially influential observations (β = 0.147, p = .014). → interaction-2

Evidence¶

interaction-1¶

Conrad Borchers, Lijin Zhang, Kexin Yang, Tomohiro Nagashima, and Benjamin W. Domingue. 2026. Understanding Student Effort Using Response-Time Propensities During Problem Solving. Proceedings of the Thirteenth ACM Conference on Learning @ Scale (L@S '26). https://doi.org/10.1145/3774398.3811624

q2 · i? · associational · r2

Linear regression on student-level parameters from the eight algebra tutoring deployments (794 students), predicting iAFM learning-rate slopes from RT propensity, prior proficiency, and their interaction. The article reports the interaction at "𝛽= 0.11" and interprets slower propensities as constructive processing for higher-proficiency students.

"Critically, there is a significant RT propensity× prior proficiency interaction (𝛽= 0.11, 95% CI[0.03, 0.18], 𝑝=. 004; Table 4), indicating that the association between RT propensity and learning rate differs by students' baseline proficiency."

interaction-2¶

Conrad Borchers, Lijin Zhang, Kexin Yang, Tomohiro Nagashima, and Benjamin W. Domingue. 2026. Understanding Student Effort Using Response-Time Propensities During Problem Solving. Proceedings of the Thirteenth ACM Conference on Learning @ Scale (L@S '26). https://doi.org/10.1145/3774398.3811624

q2 · i? · associational · r2

Sensitivity analysis re-estimating the interaction model after excluding flagged influential cases (98/787, 12.5%). The interaction "remained significant (𝛽= 0.147, p = .014)", addressing the influence of a small cluster of slow-responding students with low learning rates.

"We flagged potentially influen-tial observations (Cook's D > 4/N, leverage > 2p/N, or |studentized residual| > 3) and re-estimated the model excluding these cases (98/787, 12.5%). The RT propensity–prior proficiency interaction remained significant (𝛽= 0.147, p = .014)."

Discussion¶

Related Claims¶

• Higher prior proficiency is associated with faster step-level response-time propensities — related
• RT propensity shows no reliable main effect on learning rate, while prior proficiency is positively associated with learning rate — related
• The RT propensity–learning relationship is strongest early in the class session and attenuates later — related
• Response-time propensities show moderate to strong within-student stability across practice session slices — related
• Baseline interaction quality moderates component-outcome relations, with strongest relations for teachers with the lowest baseline quality — related
• Faster rate of learning may be negatively related to long-term retention (efficiency-effectiveness trade-off) — related