The prompting intervention did not reduce general propagation of user errors into AI advice¶

Claim · All claims
Evidence · 1 study · causal r2 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? A binomial model predicting carryover of user errors showed no significant condition×time interaction (b = 0.152, p = .712), indicating no reduction in general error propagation. → Cansu Koyuturk 2026

Evidence¶

Cansu Koyuturk 2026¶

Cansu Koyuturk, Sabrina Guidotti, and Dimitri Ognibene. (2026). The Hidden Cost of Contextual Sycophancy: an AI Literacy Intervention in Human–AI Collaboration. https://arxiv.org/abs/2605.18372 doi:10.48550/arxiv.2605.18372

q3 · i? · causal · r2

Binomial model on the 60-participant pre/post experiment testing whether the intervention changed the assistant's tendency to carry over users' incorrect inputs. The article reports "b = 0.152, SE = 0.411, z = 0.37, p = .712"; no effect size is printed, so the null is reported as a non-significant test, not equivalence.

"A binomial model predicting the carryover of user errors showed no significant condition×time interaction, b = 0.152, SE = 0.411, z = 0.37, p = .712, indicating no reduction in general error propagation."

Discussion¶

Related Claims¶

• Carryover of user errors into AI advice reduces both advice quality and final ranking accuracy — related
• The sycophancy-focused intervention significantly reduced the assistant's positional mirroring of incorrect user rankings and rank-order alignment — related