Suspecting AI cheating, Ivy League prof ordered in-person final; scores fell 50%

Original Article ↗ Hacker News Discussion ↗

At-home exams, proctoring, and cheating

  • Many see take-home testing as “structurally” cheatable, long before AI; AI just makes it obvious.
  • Several describe strict online proctoring setups (locked-down browsers, external webcams, room scans, audio monitoring) and claim these can make cheating harder than in-person exams.
  • Others argue the only robust structural fix is in-person exams; surprise at treating this as a new dilemma, since pre-COVID most exams were already in person.

What AI is doing to assessment

  • Some say mass AI use shows grades no longer measure learning, invoking Goodhart’s law: once degrees/grades are targets, they stop being good metrics.
  • A key concern: many students may not even see AI use as “cheating.”
  • One view: banning AI is like banning calculators or the internet; academia must redesign assignments to assume AI as a cognitive tool and ask harder, higher-order questions.
  • Counterview: if scores drop ~50% without AI, students didn’t just lose “memorization” help; they never learned the material.

Purpose and value of college and grades

  • Several see college primarily as a credential and a “box to check,” not a learning experience.
  • Cheating en masse undermines the signaling value of degrees; suggestions range from expulsion of cheaters to extreme proposals like criminal penalties (which others criticize).
  • Some argue most people with a pure “credential” mindset shouldn’t be in university at all; better trade and apprenticeship paths are advocated.

Ivy League, intelligence, and privilege

  • The article’s claim that Ivy students are “by definition intelligent” is widely criticized.
  • Many note Ivy cohorts are a mix of academically strong and highly privileged students; wealth and preparation often matter as much as raw ability.
  • A long subthread debates whether privilege entails greater moral obligations, spiraling into arguments over taxes, fairness, and national exceptionalism.

Future workforce and political economy

  • Some predict a generation stuck in gig work as AI erodes both the meaning of degrees and entry-level jobs, “hollowing out” society.
  • Others argue the core problem is capitalism and credentialism, not AI per se; AI simply exposes that a degree was never a reliable proxy for expertise.
  • There’s disagreement over whether AI leads toward more egalitarian systems, harsher “hyper-capitalism,” or even fascism; participants dispute how likely each is.

How education might adapt

  • Suggested reforms:
    • Shift to in-person written exams, oral exams, and practical demonstrations (e.g., one-on-one checkride-style evaluations).
    • Use personalized exams (e.g., asking students to explain their own submitted code) to reveal who really understands the work.
    • Redesign curricula so students learn fundamentals and then use AI as an amplifier, not a crutch.
  • Some educators report bimodal outcomes: a subset of students clearly learn and thrive, while many rely on AI or other shortcuts and cannot explain their own work.

Ethics, norms, and student attitudes

  • Multiple commenters report that most classmates cheated even before AI (e.g., using summaries and plagiarism), suggesting a long-standing culture problem.
  • Several argue that if institutions quietly tolerate cheating (instead of clearly punishing it), the incentive to learn collapses.
  • Others stress that education should refocus on the motivated minority who genuinely want to learn, while trying—but not assuming—to bring more students into that group.