Open ChatGPT support produces stronger self-serving attribution bias than ChatGPT support with reflection¶

Claim · All claims
Evidence · 2 studies · 2 causal r2 · q3 · i2 medium
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i2 The open ChatGPT condition showed a higher AI-specific attribution-bias index than the reflection condition (0.42 vs. 0.21, d = 0.41). → Ren S 2026 q3 i? After incorrect outcomes, attribution specifically to AI advice was higher in the open ChatGPT condition than in the reflection condition (estimated difference 0.46 points, Holm-adjusted p = 0.004). → Ren S 2026 (2)

Evidence¶

Ren S 2026¶

Ren S (2026) College students’ metacognitive awareness of generative-AI reliance: an experimental study of decision confidence and attribution bias. Front. Psychol. 17:1926110. https://doi.org/10.3389/fpsyg.2026.1926110

q3 · i2 · causal · r2

Participant-level composite self-serving attribution-bias index (internal credit after correct, external blame after incorrect outcomes) compared across the two AI-assisted conditions; "d = 0.41". The independent condition served only as a descriptive benchmark.

"The open ChatGPT condition showed a higher attribution-bias index ( M = 0.42,SD = 0.54) than the reflection condition (M = 0.21,SD = 0.48), t(222.94) = 3.10,p = 0.002,d = 0.41, 95% CI for the mean difference [0.08, 0.34]."

Ren S 2026 (2)¶

Ren S (2026) College students’ metacognitive awareness of generative-AI reliance: an experimental study of decision confidence and attribution bias. Front. Psychol. 17:1926110. https://doi.org/10.3389/fpsyg.2026.1926110

q3 · i? · causal · r2

Scoring-sensitivity analysis of raw attribution sources within the incorrect-outcome subset of the two AI-assisted conditions; the H4 conclusion did not depend on the composite formula. No standardized effect size is printed for this contrast.

"attribution specifically to AI advice was higher in the open ChatGPT condition than in the reflection condition (estimated meandifference=0.46pointsonthe1–7scale,95%CI[0.16,0.76], Holm-adjusted p = 0.004), whereas the condition contrast for task-difficultyattributionwassmaller."

Discussion¶

Related Claims¶

• Better awareness calibration co-occurs with lower incorrect-advice acceptance and lower attribution bias, interpreted associationally only — related
• Reflection improves correspondence between perceived and behavioral reliance on ChatGPT — related
• A brief metacognitive reflection prompt reduces acceptance of incorrect AI advice compared with open ChatGPT support alone — related
• Open ChatGPT support increases college students' final decision confidence relative to independent decision-making, and reflection attenuates the boost — related
• ChatGPT recommendation accuracy was comparable between the open-support and reflection conditions, ruling out unequal exposure to erroneous advice — related