Revealed: Thousands of UK university students caught cheating using AI | Higher education | The Guardian

Guardian investigation finds almost 7,000 proven cases of cheating – and experts says these are tip of the iceberg

概览

收录于

2026年3月8日

相关情境

受众

parent

年级范围

十二年级(高四)–十二年级(高四)

页面类型

Article

简介

UK University AI Cheating Investigation

  • Rising AI Misconduct: Proven cases of AI-assisted cheating in UK universities rose from 1.6 per 1,000 students in 2022-23 to 5.1 per 1,000 in 2023-24. Projections for the current year suggest a further increase to 7.5 per 1,000.
  • Decline in Traditional Plagiarism: As AI misuse has grown, traditional plagiarism rates have fallen, dropping from 19 per 1,000 students in 2022-23 to 15.2 in 2023-24.
  • The "Tip of the Iceberg": Experts warn that recorded cases significantly underrepresent the actual scale of AI cheating. A University of Reading study found that AI-generated work went undetected 94% of the time.
  • Detection Challenges: Unlike traditional plagiarism, AI-generated content is notoriously difficult to prove. Many universities (over 27% of those surveyed) do not yet track AI misuse as a distinct category of academic misconduct.
  • Student Usage: A Higher Education Policy Institute survey found that 88% of students use AI for assessments. Students report using these tools for brainstorming, structuring, and supporting learning difficulties (e.g., dyslexia), rather than just direct copying.
  • Evasion Tactics: Students are increasingly using "humanising" tools and paraphrasing software to bypass university AI detectors.
  • Sector Response:
    • Universities are struggling to adapt assessment methods to a post-AI landscape.
    • Experts suggest moving away from rote learning toward assessing skills that AI cannot easily replicate, such as communication and interpersonal skills.
    • There is a debate regarding the feasibility of returning to in-person, exam-based assessments.
  • Methodology: The Guardian surveyed 155 UK universities via Freedom of Information requests; 131 provided data.

用户评价

暂无已发布的评价,欢迎率先分享您的使用体验。