Inside the university AI cheating crisis

Assessment formats and AI use

• Many courses rely mostly on papers, projects, and presentations rather than proctored exams; some universities reduced or removed exams during Covid and never restored them.
• Others report traditional models with midterms, finals, in‑class essays, language interviews, and problem‑solving exams still dominant.
• Proposed countermeasures: handwritten in‑class essays, paper‑and‑pencil or air‑gapped lab exams, oral exams/interviews, orals at scale using AI assistance, and weighting exams more heavily to offset easy homework cheating.
• Major constraint: time and labor for oral or closely proctored assessment, especially with large classes and limited TA support.

What counts as “cheating” with AI

• Described spectrum: brainstorming topics, outlining, polishing prose, full drafting, paraphrasing tools, grammar checking, or using AI to explain readings.
• Humanities educators in the thread tend to see AI‑assisted writing as cheating; some science/technical educators are more open if the ideas and analysis are original.
• Participants highlight a large unresolved gray area and call for clearer definitions (e.g., AI‑generated then edited vs. human‑written then AI‑edited).
• One suggestion: require students to submit prompts as part of grading to expose how AI was used.

Detection tools and their limits

• Turnitin plagiarism detection is variously described as:
  • Expensive with many false positives and reliance on crude similarity metrics.
  • Still useful for catching blatant copying and paraphrasing.
• AI‑detection is widely viewed as unreliable “snake oil,” with concerns about:
  • High false‑positive rates (including non‑native speakers).
  • Lack of independent validation of accuracy.
  • Arms‑race dynamics as prompts/styles change.
• Newer tools that record the writing process (keystrokes, edits) are described; they may work now but raise evasion concerns, privacy/FERPA issues, and face adoption barriers (apathy, red tape, cynicism).

Learning, incentives, and ethics

• Some students use AI to save time or clarify material; others may bypass learning entirely.
• Debate over analogy to calculators: baseline skills are seen by some as essential to later understanding and to critiquing AI output; others argue much hand‑work is unnecessary busywork.
• Several comments criticize higher ed for emphasizing credentials, curves, and high‑stakes grading, making AI a rational way to “game” a zero‑sum system.
• Others stress personal integrity and long‑term self‑harm from cheating, while noting broader cultural distrust of institutions and role models who succeed via dishonesty.

Future of essays and assignments

• Some argue if AI can do an assignment well, the assignment design is obsolete; call to move away from formulaic essays toward presentations, more authentic tasks, or different communication forms.
• Others defend essays as a core way to develop thinking and writing, noting essay‑like writing is common outside academia (editorials, blogs, long posts).