Skip to content
Sign in
Technology · Global

Richard Borcherds says AI use has changed how he grades homework

UC Berkeley mathematician Richard Borcherds says he stopped grading homework for correctness because he believes doing so encourages students to use AI rather than work independently. He says AI use on homework is widespread and that conventional methods cannot reliably distinguish AI-generated work from students’ own work.

AI-assisted synthesis · Based on 4 outlets' reporting · Article text read from 2 of 4 outletsHow it's made

View timeline · 8 entries
4 outlets2 regionsLatest reporting 3 min read
Coverage: MixedL 0 · C 2 · R 2

What happened

Key facts

  • Borcherds said student use of AI for homework is widespread (per Breitbart, livemint.com).
  • Borcherds said grading for correctness creates an incentive to use AI and penalizes students who do their work honestly (per Breitbart).
  • Borcherds said there is no easy way to distinguish human-generated work from AI-generated work (per Breitbart).
  • Borcherds said AI sometimes made characteristic errors that could reveal its use, but that it now rarely makes obvious mistakes (per Breitbart, livemint.com).
  • Borcherds is a mathematician who won the Fields Medal in 1998 (per livemint.com).
  • Borcherds warned that if AI continues improving at its current rate, it may outperform human mathematicians in a few years (per Breitbart).
  • Borcherds said he had no advice about what to do and that ChatGPT also could not say what would happen (per Breitbart).
  • The New York Post reported on Borcherds’s comments (per livemint.com, Breitbart).

The brief

UC Berkeley mathematician Richard Borcherds says he has stopped grading homework for correctness, arguing that conventional marks can encourage students to use artificial intelligence rather than complete assignments independently.

He described student use of AI on homework as widespread and said grading for correct answers can disadvantage students who choose to do the work themselves. Borcherds said it is difficult to tell whether submitted work came from a student or an AI system. He said characteristic mistakes once offered occasional clues, but that AI now rarely makes obvious errors.

His concern, as he described it, is that correctness-based grading rewards an outcome without reliably showing who produced the work. The accounts emphasize different aspects of that decision.

Breitbart foregrounds Borcherds’s concern that grading for correctness creates an incentive to use AI and penalizes honest students, alongside his uncertainty about how the technology may develop. Livemint presents the change as reflecting a broader loss of meaning in conventional grading as AI-assisted work becomes common.

Its headline also says students score top marks, but its excerpt does not provide a figure or an example supporting that characterization. Borcherds, a mathematician who won the Fields Medal in 1998, also warned that continued improvement in AI could leave human mathematicians unable to do some work better than machines.

He framed that as a forecast conditional on AI continuing to advance at its current pace, not as a description of what has already happened. He said he did not know what advice to give and that ChatGPT could not predict what would happen either. The comments were reported by the New York Post, according to Livemint.

Borcherds’s remarks point to a practical assessment problem: when AI can produce work with few recognizable errors, correctness alone may not show whether a student understands the material or completed the assignment without assistance. His stated reason for changing grading is that the incentive attached to correct answers can push students toward AI use.

The coverage documents Borcherds’s decision and his account of why he made it, but it does not describe a replacement grading method or indicate how the change affects students’ marks. The central issue in his comments is the tension between evaluating correct answers and assessing students’ own work when AI-generated submissions are increasingly difficult to identify.

Why it matters

  • Students who complete homework independently may be disadvantaged by correctness-based grading if, as Borcherds argues, those marks incentivize others to use AI.
  • For UC Berkeley students, the difficulty of distinguishing human work from AI-generated work raises a concrete assessment problem: correct answers may not show who did the work.
  • Borcherds’s warning that AI could outperform human mathematicians if it continues improving at its current rate highlights the possible stakes for mathematics, while remaining a conditional forecast rather than an established outcome.

What to watch next

  • Whether Borcherds describes a grading approach that does not rely on correctness alone.
  • Whether Borcherds reports how the change affects students’ marks.
  • Whether Borcherds offers guidance on assessing homework as AI use continues to develop.

How it's being covered

Coverage analysisShow details
4 source articles · 4 unique outlets
Outlet political mix: Mixed· 2 Center · 2 Right
Where the outlets are based: US/Canada 75% · Asia 25%
Outlet political mix
Mixed
Left
Center
Right
0 outlets
2 outlets
2 outlets
The marker shows coverage balance: each unique outlet counts once.
Geography
Where the outlets covering this story are based.
4 unique outletsDominant coverage: US/Canada
Coverage imbalance
  • Coverage is currently concentrated among outlets based in US/Canada (3 of 4 geographically-identified sources).
How we measure coverage

How outlets' accounts compare

2 framing differences · 2 uncorroborated claims · 2 information gaps
The reports repeat the same account with different emphasis; its central claims are not independently confirmed in the reporting Updat3 read.Compared across 2 reports whose article text Updat3 read; 2 other outlets were checked by headline only.
Framing / emphasis differences
Sources emphasise different angles, actors or consequences of the same reporting.
  • Breitbart highlights Borcherds’s view that correctness-based grading incentivizes AI use and penalizes honest students; Livemint frames the change as conventional grading becoming less meaningful as AI use grows.
  • Livemint’s headline says students score top marks, while the excerpt does not provide a score, number, or example.
Updat3: These are differences of emphasis, not disputes about what is reported. They do not show that the reported claims are true.
One-sided or uncorroborated claims
Claims that come from one party or one source family and are not independently corroborated.
  • Borcherds said grading for correctness can create an incentive to use AI and disadvantage students who work honestly.
  • Borcherds warned that AI could outperform human mathematicians if it keeps improving at its current rate.
Updat3: Not independently corroborated yet — treat as a claim, not an established fact.
Not established by the reporting
Information the available reporting does not contain or confirm. This is a gap, not a disagreement.
  • The sources do not identify a replacement grading method or explain how the change affects students’ marks.
  • The sources do not describe how Borcherds grades homework after stopping correctness-based grading.
Updat3: Updat3 cannot conclude anything about this from the current sources.

Timeline

Developments from this story's sources, dated background we compiled and fact-checked, and our earlier coverage. Each row says which.
  1. Borcherds is a mathematician who won the Fields Medal in 1998.
    Reported in this story's sourcesSource:livemint.com
  2. Sam Altman, Greg Brockman and other co-founders launched OpenAI as a research organization.
    Background · compiled and fact-checked by Updat3
  3. In 2017, Google researchers introduced the Transformer architecture in the paper “Attention Is All You Need,” describing a neural-network design that became central to later language models.
    Background · from our historical context
  4. OpenAI published a paper introducing its first GPT model.
    Background · compiled and fact-checked by Updat3
  5. OpenAI released ChatGPT on November 30, 2022, making a conversational generative-AI tool widely accessible to students and teachers.
    Background · from our historical context
  6. New York City’s Department of Education announced a restriction on ChatGPT on its networks and devices.
    Background · compiled and fact-checked by Updat3
  7. OpenAI released GPT-4 on March 14, 2023, extending the capabilities of the model family behind ChatGPT.
    Background · from our historical context
  8. New York City’s Department of Education lifted its restriction on ChatGPT.
    Background · compiled and fact-checked by Updat3

How we got here

OpenAI released ChatGPT on November 30, 2022, making a conversational generative-AI tool widely accessible to students and teachers. On January 12, 2023, the New York City Department of Education announced a ban on ChatGPT on its networks and devices, citing concerns that included academic integrity; the department lifted the restriction on May 18, 2023.

OpenAI released GPT-4 on March 14, 2023, extending the capabilities of the model family behind ChatGPT. In 2017, Google researchers introduced the Transformer architecture in the paper “Attention Is All You Need,” describing a neural-network design that became central to later language models.

Original sources

4 articles from 4 outlets

Every story links to the reporting it was built from. How we choose sources

Spotted a factual error, a broken link or a misleading summary? Tell us.