Neither self-efficacy nor course type statistically predicted students' decisions to continue ChatGPT feedback conversations¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Self-efficacy scores did not predict continuing the conversation with ChatGPT (z(277) = -1.125, Pr(>|z|)=0.260). → Sperber 2025 q2 i? Continuation decisions did not differ statistically between large WI courses and small writing classes (χ2(1) = 0.701, p = 0.402). → Sperber 2025 (2)

Evidence¶

Sperber 2025¶

Sperber, Lisa; MacArthur, Marit; Minnillo, Sophia; Stillman, Nicholas; Whithaus, Carl. (2025). Peer and AI Review + Reflection (PAIRR): A human-centered approach to formative assessment. Computers and Composition 76. https://doi.org/10.1016/j.compcom.2025.102921

q2 · i? · design · r2

Binary logistic regression in R (glm function) on spring survey data testing whether self-efficacy scores ranging from -2 to +2 predicted continuing the conversation; the test was not significant and no effect size is printed.

"self-efficacy scores did not predict students’ decision to continue the conversation [z(277) = -1.125, Pr(>|z|)=0.260"

Sperber 2025 (2)¶

Sperber, Lisa; MacArthur, Marit; Minnillo, Sophia; Stillman, Nicholas; Whithaus, Carl. (2025). Peer and AI Review + Reflection (PAIRR): A human-centered approach to formative assessment. Computers and Composition 76. https://doi.org/10.1016/j.compcom.2025.102921

q2 · i? · design · r2

Chi-squared test in R (infer package chisq_test) comparing continuation decisions between small writing classes and large WI classes; the authors describe the result as surprising and no effect size is printed.

"student feedback preferences were not statistically different in large WI courses compared to small writing classes [χ2(1) = 0.701, p = 0.402"

Discussion¶

Related Claims¶

• Students who continued feedback conversations with ChatGPT perceived more utility in AI feedback than one-shot users — related
• Structural paths are stable when continued-use intention and current use are modeled as separate outcomes, and AI anxiety is not directly related to continued use — related
• Self-efficacy, but not self-reported strategy use, predicts subsequent exam grades in the university course — related
• Voluntary classroom uptake of teacher-assigned AI discussions was high on average (85.0% classroom participation) but heterogeneous across teachers and discussions — related