Source provenance labels functioned as a seal of authority: 11 of 12 participants failed to detect a planted source-card mismatch¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r1–r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? In a controlled deception experiment, only 1 of 12 participants detected the source-card mismatch; the remaining 11 accepted it without hesitation. → Ruqi Sun 2026 q2 i? Perceived trust trended higher under VeriForge (M=5.92 vs M=4.75) but did not reach significance (p=.125), while perceived authorial autonomy was identical (M=6.17 vs M=6.17; p=1.0). → Ruqi Sun 2026 (2)

Evidence¶

Ruqi Sun 2026¶

Ruqi Sun, Jiaping Li, Wenhui Tao, Ximing Zheng, Yuefeng Tan, Jiahao Wei, and Yuxin Ma. (2026). VeriForge: Mitigating Latent Knowledge Gaps in Narrative Drafting via Mixed-Initiative Scaffolding. https://arxiv.org/abs/2608.09698 doi:10.48550/arxiv.2608.09698

q2 · i? · causal · r2

Controlled deception in the within-subjects study: the source passage on the second Knowledge Card was replaced with an unrelated passage while keeping the AI-generated description unchanged, creating a detectable mismatch. The test measured evidence inspection, not hallucination detection.

"Only 1 of 12 participants detected the source-card mismatch. The remaining 11 did not notice."

Ruqi Sun 2026 (2)¶

Ruqi Sun, Jiaping Li, Wenhui Tao, Ximing Zheng, Yuefeng Tan, Jiahao Wei, and Yuxin Ma. (2026). VeriForge: Mitigating Latent Knowledge Gaps in Narrative Drafting via Mixed-Initiative Scaffolding. https://arxiv.org/abs/2608.09698

q2 · i? · causal · r1

Exploratory RQ3 survey measures excluded from FDR correction. The authors attribute the compressed trust difference to both conditions sharing the same source-attributed retrieval backend.

"Perceived trust trended higher under VeriForge (𝑀= 5.92vs. 𝑀= 4.75) but did not reach significance (𝑝=. 125)"

Discussion¶

Related Claims¶

• Surface validation can function while substantive validation fails: most participants detected flawed AI recommendations yet continued following them — related
• Experimental-group students detected AI errors and enacted verification and ethical self-regulation as routine practices — related
• Student experiences with GenAI in design thinking fall into four themes: perceived benefits, ethical concerns, hesitance and acceptance, and critical validation — related
• Perfect AI execution cannot remove four limits: accuracy does not settle end-setting, safety does not confer legitimacy, detection does not secure integrity, and outputs do not replace participation — a broader claim this one bears on
• VeriForge raises perceived creativity support and lowers workload relative to a same-retrieval baseline, with mixed survival after correction — related