📄 Full text: arXiv:2605.05472 · local
Definition
An instructional approach that deliberately leverages AI errors, hallucinations, and limitations as teaching tools to foster higher-order thinking. Rather than viewing AI mistakes as failures to be avoided, this pedagogy treats them as cognitive provocations that demand analysis, evaluation, and reflection from students. Proposed by Hosseini (2026) in a database design course context.Mechanism
Students interact with AI-generated outputs that contain intentional or known errors. They must: 1. Analyze the output for correctness against disciplinary standards 2. Evaluate where and why the AI went wrong 3. Reflect on what the error reveals about both the domain and AI limitationsThis maps directly to the upper levels of Bloom's taxonomy (Analyze, Evaluate, Create) and engages metacognitive processes central to metacognition.
Relationship to Existing Approaches
- Complements socratic-ai-dialogue: while Socratic approaches use questions to guide reasoning, mistake pedagogy uses erroneous outputs as the provocation
- Extends ai-literacy: students learn not just to use AI but to critically evaluate its outputs
- Addresses the llm-fallacy-misattribution problem by making AI's limitations visible and discussable
- Contrasts with tutoring-specific-vs-general-ai: here the AI's imperfection is the feature, not the bug
Open Questions
- Does mistake-based pedagogy transfer across disciplines beyond STEM?
- What is the optimal error difficulty — too obvious vs. too subtle?
- How does this approach affect trust in AI tools long-term?
Related Pages
- institutional-change-framework-ai — Six-dimension framework for adapting institutional change models in STEM to generative AI
- conversational-ai-tutors-framework — Affect detection and knowledge tracing as proven methods to keep in AI tutors
- moodle-ai-tutoring-deep-learning — Complements AI-mistake pedagogy with deep understanding scaffolding
- sequenced-ai-feedback-learning — Cao et al. RCT: direct corrective feedback outperformed encouraging scaffolded feedback for learning outcomes
- ai-generated-slides-student-perception — Student bias against perceived AI-generated content
- genai-performance-vs-learning — AI errors as learning opportunities