人工智慧檢測出數千篇可疑的癌症研究論文
AI Detects Thousands of Suspicious Cancer Research Papers
Updated at: July 17, 2026 at 11:00 AM
近期發表於《英國醫學期刊》(The BMJ) 的一項研究，揭露了醫學進步面臨的一項日益嚴峻的威脅：虛假研究的興起。
A recent study published in The BMJ has shed light on a growing threat to medical progress: the rise of fraudulent research.
由亞德里安·巴內特 (Adrian Barnett) 教授帶領的研究團隊，分析了 1999 年至 2024 年間發表的 260 萬篇癌症研究論文。
Led by Professor Adrian Barnett, researchers analyzed 2.6 million cancer studies published between 1999 and 2024.
透過一種客製化的機器學習工具，他們辨識出超過 25 萬篇呈現「論文工廠」(paper mills) 特徵的論文，這些組織專門販售量產的偽造科學手稿。
Using a custom machine-learning tool, they identified over 250,000 papers that display characteristics of 'paper mills'—organizations that sell mass-produced, fake scientific manuscripts.
問題的嚴重性令人震驚。
The scale of the issue is alarming.
可疑研究的比例從 2000 年代初期的總數 1% 左右，飆升至 2022 年的 16% 以上。
Suspicious research has surged from just 1% of the total in the early 2000s to over 16% by 2022.
該人工智慧 (AI) 採用 BERT 模型，透過掃描標題和摘要中常見的詐欺性工作「指紋」(linguistic fingerprints)，達到了 91% 的準確率。
The AI, which uses a BERT-based model, achieved 91% accuracy by scanning titles and abstracts for linguistic 'fingerprints' common in fraudulent work.
儘管這項技術能作為「科學垃圾郵件篩選器」，協助期刊進行編輯審查，但研究結果凸顯了重大挑戰。
While this technology acts as a 'scientific spam filter' to assist journals during editorial review, the findings highlight significant challenges.
發表論文的高壓以及有限的同儕審查能力，創造了讓偽造研究得以滋長的環境。
High pressure to publish and limited peer-review capacity have created an environment where fake research thrives.
確保醫學研究的完整性，對於有效的經費分配、科學進步，以及維護公眾對全球衛生機構的信任至關重要。
Ensuring the integrity of medical research is essential for effective funding, scientific progress, and maintaining public trust in global health institutions.
