Responsible AI Use in Research

The rapid adoption of AI in research has outpaced policy development, but a consensus is forming. Key principles include: Authorship exclusion — AI systems cannot meet ICMJE criteria (they cannot be accountable), so they cannot be listed as authors. Transparency — researchers must disclose when and how AI tools were used, typically in the Methods section or an acknowledgments statement. Accountability — the human authors bear full responsibility for accuracy, originality, and integrity of all content, including any AI-assisted portions. Verification — AI-generated content must be verified against primary sources. LLMs can 'hallucinate' — generating plausible-sounding but fabricated references, statistics, or claims. Specific risks in research include: fabricated citations (AI may generate references that look real but do not exist), confident but incorrect statistical interpretations, paraphrasing that inadvertently plagiarizes source material, and generation of plausible but false data patterns. Responsible use means treating AI as an assistant, not an authority. Use it for: brainstorming, improving grammar and clarity, generating code scaffolds, and summarizing large texts. Always verify factual claims, check every cited reference exists, and never present AI output as your own expert judgment without verification.