Trinity College philosophy's AI ban sparks debate over enforceability and alternatives
Trinity College Dublin's philosophy department has announced a blanket ban on student use of AI in coursework, sparking ongoing debate. Critics call it "a crazy policy that will never work," while supporters counter that the ban is not unenforceable and have proposed several institutional designs. The core of the dispute is the definition of "enforceability" and the standard of proof, not whether AI should be restricted at all.
Confirmed
- Trinity's philosophy department has rolled out a comprehensive AI ban, with detection relying on tools like Pangram—described as reliable but still controversial (per @dioscuri).
- @rbnmckenna86 offers the central defense: strict policies are easier to enforce than lenient ones—any evidence of AI use constitutes a violation, so a perfect detection tool isn't required; oral exams (vivas) can serve as a verification mechanism.
- @d08890 points out critics' logical error: swapping "enforceable" for "perfectly or near-perfectly enforceable"; in reality, no academic integrity policy is perfectly enforceable, so rejecting strict bans on grounds of "imperfection" doesn't hold.
- Alternatives were made concrete: @rbnmckenna86 suggests dropping the "beyond reasonable doubt" standard in favor of the lower "preponderance of evidence" standard, with evidence gathered and assessed against the standard by someone outside the course; @d08890 proposes tiered proof by penalty severity—minor sanctions like a minimum passing grade could use a lower evidentiary bar, while severe ones like expulsion would require a higher bar, still gameable but much harder to exploit.
- @dioscuri remains pessimistic: without hard evidence, students can always deny, so the ban only catches the careless; professors get pushed into the "bad guy" role with no incentive to enforce, and apart from a few AI opponents, most people accept the open secret that "everyone is using it."
Unconfirmed
- The actual reliability and false-positive rate of the Pangram detection tool remain disputed in the discussion, with neither side reaching a conclusion.
- Whether the "preponderance of evidence" standard or tiered proof schemes will actually be adopted by any institution is not addressed in the material.
Why it matters
This debate reflects the fundamental dilemma of academic integrity policy in the AI era: traditional plagiarism-detection logic doesn't apply to generative AI, and standards of proof, penalty gradients, and enforcement costs all need redesign. Supporters' institutional proposals offer a rebuttal framework to the popular "bans are unenforceable" narrative, while opponents' questions about professors' enforcement incentives are a reminder that any policy will become a dead letter if frontline teachers don't buy in.
2026-11-17 ~ 2026-11-17 · 6 related posts
Primary sources
- Trinity College philosophy dept bans AI — and a strict ban may be easier to enforce — rbnmckenna86 ·
- A weaker 'preponderance of evidence' standard could make AI bans enforceable — rbnmckenna86 ·
- Why AI coursework bans fail: professors lack incentives, only assessment redesign or Pangram-style judges remain — dioscuri ·
- [source] Trinity College philosophy dept bans AI — and a strict ban may be easier to enforce — rbnmckenna86 · 2026-11-17
- No academic integrity policy is perfectly enforceable — the objection is a strawman — d08890 · 2026-11-17
- Without a smoking gun students can always deny AI use — bans only catch the careless — dioscuri · 2026-11-17
- [source] A weaker 'preponderance of evidence' standard could make AI bans enforceable — rbnmckenna86 · 2026-11-17
- Debating AI coursework bans: tiered evidence standards, viva checks and version history as alternatives — d08890 · 2026-11-17
- [source] Why AI coursework bans fail: professors lack incentives, only assessment redesign or Pangram-style judges remain — dioscuri · 2026-11-17