AI检测误判学术写作辅助，真实使用反而更易被罚
Why AI Detection Fails for Academic Integrity
- 用真实论文摘要测试，轻度修改也常被误标为作弊
- 未改原文2023-2025年作品被误标9%-15%，非理工类更高
- 伪装成人类写法后96%可逃过检测，建议勿单凭分数定责
机构依赖商业AI检测工具维护学术诚信，但现有工具无法区分AI编辑与完整生成内容，可能将两者均视为违规。我们在四个领域对2013–2015年与2023–2025年的已发表英文摘要进行受控研究，以tau=0.50为阈值量化政策失效。仅“精炼摘要”类轻度编辑（代表合规使用AI）被标记率高达38%至80%；未修改的2023–2025年原文被误标率为9%至15%，非理工类显著高于理工类（p<0.001）；误标率与长文本长度及学术词汇密度相关，非仅由作者意图决定。经不可检测的AI人类化处理后，逃逸率达近96%：少于4%的AI重写仍被标记（检测率<4%；假阴性率>96%）。诚实使用AI编辑的风险反而高于通过人类化伪装逃避检测。因此，检测得分不应作为单独的违规证据。
原文摘要 · Abstract (English)
Institutions use commercial AI detectors for academic integrity, yet detectors cannot distinguish AI editing from full LLM drafts and may treat both as misconduct. In a controlled study of published English abstracts (four domains; 2013 to 2015 vs. 2023 to 2025), we quantify this policy failure under proxy human/AI labels at tau=0.50. Light "refine abstract only" edits, a proxy for guideline-compliant AI assistance, are flagged at 38 to 80%. Unmodified 2023 to 2025 originals are flagged at 9 to 15%, with non-STEM rates far above STEM (p<0.001); elevated scores track long-token and Academic Word List density, not authorship intent alone. After Undetectable AI humanization, evasion is near-total: fewer than 4% of AI-labeled rewrites remain flagged (post-humanization detection rate <4%; FNR >96%). Honest AI-editing results in a higher sanction risk than humanizer-assisted evasion. Therefore, detector scores should not serve as standalone misconduct evidence.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。