Syntes: Steenbergen-Hu & Cooper (2014) — tutoring systems in college: ahead of other instruction, not clearly behind human tutors

Källa: Steenbergen-Hu, S., & Cooper, H. (2014). A meta-analysis of the effectiveness of intelligent tutoring systems on college students' academic learning. Journal of Educational Psychology, 106(2), 331–347. https://doi.org/10.1037/a0034752

Verified numbers (full text read 2026-10-10, Duke repository copy)

• Scope: 35 reports, 39 studies, 22 types of tutoring system (most often AutoTutor, ALEKS, eXtended Tutor-Expert System, WIST) in college courses.
• Overall: g = 0.32 to 0.37 against all other conditions combined; "instruction learning" (traditional plus computerised instruction) g = 0.37 (k = 36).
• Table 2, by comparison condition: human tutoring k = 3, g = −0.25 (SE 0.24, 95 % CI −0.72 to 0.22, p = .302; same fixed and random); reading computerised materials 0.22; computer-assisted instruction 0.35; traditional classroom instruction 0.37; reading printed text 0.47; self-reliant learning 0.82; no-treatment control 0.90.
• The three human-tutor studies: Reif & Scott 1999; Stankov, Glavinić & Grubišić 2004; VanLehn et al. 2007 (Study 1).
• Authors' reading: the result "was not statistically significant" and rests on three studies; averaging VanLehn 2011's step-based and substep-based human-tutor comparisons gives d = −0.05, so the findings are "largely consistent" with VanLehn 2011; tutoring systems "are not yet as effective as human tutors".

Role for C5

Supports (kill-condition test survived, 2026-10-10): the only pooled direct comparison with human tutors in this meta-analysis does not show tutoring systems clearly short of human tutors (−0.25, not significant, below the 0.3 SD threshold). Thin: three studies, wide interval, all system types pooled, college only, one study overlaps with VanLehn 2011.