In a randomised experiment, AI access raised students' enjoyment of learning (p = 0.063) with no statistically detectable effect on self-assessed knowledge or felt effectiveness¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r3 · q4
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person

Subclaims¶

q4 i? Treated students report enjoyment of the learning task 0.58 points (11 percent) higher on a 0–10 scale, while self-assessed knowledge and feeling skilled or effective do not change significantly. → Zara Contractor and Germán Reyes 2026 q4 i? Self-assessed knowledge rises sharply in both groups after learning, leaving an ITT of −0.10 points that is statistically indistinguishable from zero. → Zara Contractor and Germán Reyes 2026 (2)

Evidence¶

Zara Contractor and Germán Reyes 2026¶

Zara Contractor and Germán Reyes. (2026). Experimental Evidence on the Learning Impact of Generative AI. https://arxiv.org/abs/2607.08849 doi:10.48550/arxiv.2607.08849

q4 · i? · causal · r3

Post-learning survey measure from the randomized experiment; the enjoyment effect is marginal (p = 0.063), which the authors describe as AI making learning more enjoyable. The article also reports the share of students rating enjoyment at or above the sample median rises by 12.1 pp (p= 0.082), while the effect on feeling skilled or effective is 0.06 points (p= 0.834).

"Enjoyment of the learning task rises by ˆβ= 0.58points on the 0–10 scale (p= 0.063), an 11 percent improvement over the control mean of 5.22."

Zara Contractor and Germán Reyes 2026 (2)¶

Zara Contractor and Germán Reyes. (2026). Experimental Evidence on the Learning Impact of Generative AI. https://arxiv.org/abs/2607.08849

q4 · i? · causal · r3

Null result on the self-assessed knowledge outcome (0–10 scale) from Session One of the randomized experiment; objective test performance in the same session does respond to AI access. A similar null appears in Session Two (ITT 0.22 points, p= 0.250).

"Self-assessed knowledge in the control group rises sharply, from 1.56 to 4.79 on the 0–10 scale (p <0.001). The rise is nearly identical among treated students, leaving an ITT ofˆβ=−0.10points, which is statistically indistinguishable from zero."

Discussion¶

Learner Variables¶

• Interest — outcome: instruction changes it
• Achievement Emotions — outcome: instruction changes it

Related Claims¶

• Mastery learning has overall positive effects on student affect, with the exception of slightly lower grade expectations — related
• Cignition group tutoring produced slightly higher math confidence and enjoyment than control in an RCT — related
• Teachers who find using AI enjoyable, engaging and satisfying (intrinsic/interest value) may be more motivated to use it (theoretical argument). — related
• Access to a generative-AI assistant raises task performance for both lower- and higher-education adults, with significantly larger gains for lower-education participants — related
• Students report higher engagement and motivation with the AI tutor, with comparable enjoyment and growth mindset — related
• SupplyNet was rated significantly more immersive, motivating, and enjoyable than the baseline, with engagement showing the most pronounced difference. — related
• Most reviewed studies reported significantly positive AI effects on STEM learning performance, with two null results — related
• Intervention students reported significantly higher exam confidence than controls (M=4.3 vs 3.4 on a 5-point scale, d=0.76) — related
• After a guided-inquiry lesson with the collision-carts simulation, students' self-reported knowledge of the physics increased and most rated the lesson enjoyable and valuable — related
• In Study 1, ChatTutor produced significantly higher trust, enjoyment, and behavioral intentions than ChatGPT, while students were blind to condition — related
• Mental simulation training increased children's reading motivation in both grades, while control-group motivation showed no significant change — related