Apply four operational criteria—fidelity, non-augmentation, traceability, attestation—to identify transcription-only AI use¶

Strategy · All strategies
Evidence · no claims cited

Description¶

The paper proposes four operational criteria as a structured evidential framework for misconduct panels assessing whether tool use was transcription-only: fidelity (output verifiably equivalent to the handwritten original, with preserved errors as strong evidence of faithful transcription), non-augmentation (the tool must not contribute what should be written), traceability (contemporaneous artefacts including originals, images, prompt histories and outputs), and attestation (the student attests they checked the transcription and accepted no substantive suggestions). The criteria are designed to be conservative and applicable by non-specialist decision-makers.

Design Implications¶

Context¶

Requirements¶

• Decision-makers require testable conditions applicable to a specific instance of tool use; without such criteria the function-not-platform principle remains procedurally unusable

Constraints¶

• The criteria are a proposal and have not been empirically tested against a corpus of cases or in a panel-member decision study
• The criteria are not proof against all possible forms of deception; logs can be incomplete, selectively presented, or fabricated
• Whether the framework operates fairly when applied by non-specialist panel members is an empirical question

Target Learners¶

• higher education students
• academic misconduct panel members

Target Learning Goals¶

• procedural fairness in misconduct proceedings
• evidential standards for AI-use allegations

Related Strategies¶

• Institutions should define generative-AI prohibitions by function, distinguish generation from format conversion, and require validated evidential methods before sanctions
• Replace detector-centred assessment with a four-stage process-oriented model: declared starting point, versioned drafting, process artefacts, and targeted oral verification
• Assess GenAI misconduct evidence against three probative credentials — relevance, credibility and inferential force — before imposing sanctions

Examples¶

-

Key Sources¶

• Craig Wright. (2026). Transcription is not generation: distinguishing non-generative AI tool use from academic misconduct in higher education assessment. International Journal for Educational Integrity. https://doi.org/10.1007/s40979-026-00234-w