Science & Tech · AI · 8 hrs ago
AI expert urges exam rules based on what tests are meant to measure
UCL professor Rose Luckin says exam rules should depend on what is being tested.
She made the comments in an interview about recent cheating concerns in South Korea.
Luckin said secretly using AI in a closed exam to answer questions is cheating when the test is meant to assess a candidate’s own knowledge and judgment.
She said AI should not be banned in every exam, because some tests may assess how well people use and check AI.
She proposed telling candidates in advance whether AI is allowed, requiring them to disclose its use and checking their understanding.
Luckin recommends combining basic tests with AI-supported tasks and oral or practical checks.
These methods could help confirm that candidates understand their answers and can spot errors in AI output.
She says exam bodies should update questions and share information as AI changes, while supervised written exams can still play a role.
Rose Luckin, an honorary professor at University College London, says exam rules on artificial intelligence should depend on what each test is designed to assess.
She said using AI secretly in a closed exam to obtain answers is cheating.
Luckin proposed clear rules on whether AI is allowed and whether its use must be disclosed.
She recommended combining tests of basic knowledge with AI-assisted tasks and oral or practical assessments.
She said stronger supervision and detection alone cannot prevent cheating as devices become smaller and cheaper.
- Who
- Rose Luckin, honorary professor at University College London.
- What
- She urged exam bodies to set AI rules according to what tests measure and to use multiple forms of assessment.
- When
- She gave the interview on November 22; the article was published that day.
- Where
- The interview was with Yonhap News Agency in South Korea.
- Why
- She said exam rules should assess candidates’ own knowledge and judgement while accounting for AI’s ability to generate answers.
This story does not have two clearly opposing sides.
A person identified in the article as the top tax-accountant exam candidate appeared at a Seoul police office as a suspect in an AI cheating case.
Luckin gave Yonhap a written interview about AI use and exam rules.
- Interviewee
- Rose Luckin, honorary professor at University College London
- Interview date
- November 22, 2026
- Suggested approach
- Combine supervised tests of basic knowledge, AI-permitted applied tasks and oral or practical checks
- Proposed AI-use criteria
- Give candidates advance notice, require transparent disclosure and check their own understanding and judgement
What people are saying
Using concealed equipment deliberately to obtain answers in a closed exam is cheating by any definition.
Oral exams or short follow-up interviews should ask candidates to show how they arrived at an answer, what alternatives they considered and why they ruled out other alternatives.
Another option is to present an AI answer containing errors and ask candidates to find and correct them, then explain why.