随着人工智能深度融入科研工作，传统的学术规范体系正面临前所未有的挑战。一线科研人员通过具体案例指出，AI在科研全过程中的广泛应用带来了一系列新的诚信问题，急需建立与时俱进的规范体系。1
在科研创意来源上，虽然AI能够生成科研点子，但这只是初步想法，真正的原创归属应当根据谁完成了实质性创造性工作来判定。1数据处理阶段则出现了隐形造假现象——科研人员利用AI进行"美化"处理，通过去噪、补全缺失、平滑跳变等手段使数据失真，这已成为常见问题。1在验证环节，AI预测功能带来的风险同样不容忽视，因为AI预测不能替代必要的实验验证，而科学进步往往来自异常现象的发现。1
信息安全威胁也日益凸显。2025年7月，有关部门披露了一起科研人员将核心数据上传至AI应用软件导致涉密信息泄露的案件。1同年9月，ChatGPT Deep Research功能曝出零点击漏洞，攻击者可通过恶意邮件窃取敏感数据，这表明AI工具本身也存在安全隐患。1
为应对这些挑战，专家建议论文披露制度进行改革。现行的"使用/未使用"二元声明应转向分级披露模式，使披露方式与AI对研究的实际影响程度相匹配。1
As artificial intelligence becomes increasingly integrated into research workflows, scientists and academic professionals are grappling with novel questions about research integrity and publication standards. Through interviews with five frontline researchers, the discussion reveals pressing concerns that demand updated academic norms to address AI's deep involvement across all stages of scientific work.1
One of the central issues concerns the attribution of research ideas. While AI can generate scientific concepts, true originality should be determined by who completes substantive creative work rather than merely who initiated the idea.1 Beyond conceptualization, researchers face a widespread but often undetected form of data manipulation: AI "beautification" of datasets through noise reduction, gap filling, and smoothing anomalies—techniques that distort data integrity while remaining difficult to detect.1 Furthermore, AI predictions should not substitute for necessary experimental validation, as scientific breakthroughs frequently emerge from unexpected findings that predictive models might overlook.1
The risks extend beyond methodological concerns to information security. In July 2025, China's security authorities disclosed an incident where a researcher uploaded core data to an AI application, resulting in the leakage of classified information.1 Additionally, in September 2025, a zero-click vulnerability was discovered in ChatGPT's Deep Research function, which could allow attackers to steal sensitive data through malicious emails.1
To address these multifaceted challenges, experts call for modernized disclosure practices in academic publishing. Current binary declarations of whether AI was used should evolve into tiered disclosure systems that reflect the actual extent of AI's impact on research findings.1
评论
还没有评论，欢迎留下第一条。