Case 03: AI-Generated Submission Failing Authorship Interview
Student submitted a mostly AI-generated assignment and could not explain the key arguments during an authorship interview.
Evidence considered
- The submission itself, including internal inconsistencies and references the student could not locate when asked.
- An authorship interview in which the student was unable to explain the essay’s central arguments, define its key terms, or describe how its sources were selected.
- The absence of drafting history or notes.
Reasoning
The framework resolves against the student at every step: substantive AI generation was not permitted for this task; it was not disclosed; AI replaced rather than assisted the student’s work; the student evidently did not verify the output (they could not account for its references); authorship was not preserved — the interview showed the student cannot explain or defend the submitted arguments; and presenting the work as their own constitutes deception. The decisive evidence is the authorship interview, not AI-detection scores: the student’s inability to stand behind the work is established directly.
Decision
Misconduct. The submission misrepresents authorship of substantially machine-generated work in a task where that was prohibited.
Alternative interpretation
A student with genuine knowledge might interview poorly through anxiety or language barriers. That is why the interview must probe understanding supportively and why the decision rests on the whole pattern of evidence — unexplainable arguments and unlocatable references and no drafting trail — rather than interview performance alone. Here the combined evidence leaves little room for an innocent reading.
Confidence
High on the combined evidence; any single strand alone would warrant lower confidence.
Procedural fairness considerations
- The allegation and evidence must be put to the student before any decision, with an opportunity to respond and to bring support to the interview.
- Interview questions should test understanding of the submitted work, not general ability, and should account for anxiety and language background.
- AI-detector output, if any, should be treated as an unreliable prompt for inquiry — never as proof.
- The outcome and appeal rights must be communicated clearly.