Students describe concrete verification behaviors for GenAI explanations, treating evaluation as an iterative filter rather than a single accept-or-reject judgment¶

Claim · All claims
Evidence · 1 study · qualitative r3 · q2
Currency · no dated study recorded · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Participants cross-referenced AI output against course notes, conditioned trust on provenance, and probed with follow-up questions for inconsistency. → verification

Evidence¶

verification¶

Seth Bernstein and Naaz Sibia. 2026. Flawed but Memorable: Student Critical Reception of Interest-Personalized GenAI Analogies in Computing Education. Koli Calling '26. https://arxiv.org/abs/2609.06095 doi:10.48550/arxiv.2609.06095

q2 · i? · qualitative · r3

Interview and think-aloud accounts in the ten-participant study. The article reports "concrete verification behaviors rather than passive acceptance," including P4's habit of fact-checking against course notes and P6 testing analogies against worked examples.

"Participants also described concrete verification behaviors rather than passive acceptance, which speaks to how these criteria were applied in practice rather than only in the abstract."

Discussion¶

Related Claims¶

• High-achieving students describe selective AI use for clarification, summarisation, and workflow support while verifying outputs and keeping them subordinate to their own judgment — a broader claim this one bears on
• GenAI use involves substantial hidden labor of prompting, verification, and correction that remains invisible in productivity evaluation — related
• Practitioners hold near-universal commitments to human oversight and critical evaluation of AI outputs — related
• Reviewed literature reports risks of overreliance, AI errors in complex tasks, and students' uncritical acceptance of AI-generated outputs — related
• In interviews with 25 engineering students at one Swedish university, students described GenAI as acceptable when it stimulates reflection and directly copying its outputs as cheating, and one used its fallibility as a prompt to double-check — related
• Students exercise ethical agency in GenAI use through human-AI collaboration strategies, personal ethical frameworks, strategic calibration of use, and resistance to over-reliance — related
• Learner agency in GenAI-mediated learning is enacted through prompting, questioning, verification, comparison, revision, and disciplinary judgment — a broader claim this one bears on
• Trust in personalized analogies moved in no single direction, while critical scrutiny tracked domain expertise more consistently than personal interest — related
• Verification behavior lagged rising AI use: minimal or conditional verification more than doubled (5 to 12) over the semester — related
• Boundary conditions on verification and reliance include prior knowledge, learner characteristics, task stakes, verifiability, costs, accountability, AI system configuration, and multidimensional AI literacy — related
• Student narratives complicated the educator assumption that copied AI answers imply no learning, revealing invisible post-AI learning activities — related
• Students described the verification label as a reason to engage with AI-generated material that did not end their own scrutiny of it — related
• AI intuition emphasizes epistemic judgment rather than output optimization, distinguishing it from prompt engineering — related
• Learners selectively filtered and adopted AI-generated feedback while maintaining pedagogical agency — related
• A student maintained critical distance from GenAI outputs through a paradoxical trust in their untrustworthiness, deriving practical value while staying sceptical
• Non-expert students exhibit intuitive behaviors towards LLM-based AI systems when prompt engineering — related
• Active engagement with AI outputs (cross-verifying, refining prompts, revising) supports productive learning and depends on learners' prior knowledge — related
• Students' AI workflows shifted toward guided, step-by-step use and more sophisticated prompting and critical evaluation over the semester — related
• Students showed bounded trust and uneven verification behavior: 23 of 837 student turns (about 2.7%) contained explicit risk-aware moves — related
• Students use self-initiated strategies of verification, skill improvement, post-editing, pre-editing and reliance management to cope with GenAI challenges — a broader claim this one bears on
• Oral and explanation-based evaluation of AI-leveraged projects is complicated because students may break down on counterfactual questions even when they did the work — related
• The verification threshold trades hallucination reduction against over-refusal, with the default tau of 0.20 providing the best balance — related
• Instructor verification of AI-generated content before delivery is feasible and crucial for quality, per prior systems reviewed — related