دلالت‌های سیاستی در کاربست افزارهای هوش مصنوعی زایا در نگارش و داوری مقالات علمی

The rapid adoption of generative artificial intelligence (GenAI) tools, including large language models and chatbots like ChatGPT, has profoundly reshaped scientific content production. These technologies offer powerful new capabilities for researchers while raising significant concerns about the integrity, quality, and trustworthiness of scholarly publishing. This study examines the applications, benefits, risks, and policy considerations of GenAI in both scientific writing and peer review, with a particular emphasis on Iranian scholarly journals.
Researchers conducted a systematic literature review following the PRISMA guidelines, combined with qualitative content analysis of official policies from major global publishers such as Science, Elsevier, Springer Nature, IEEE, and Emerald. Data sources included peer-reviewed articles published between 2020 and 2025 in Scopus and IEEE Xplore, along with the latest publisher guidelines available as of September 2025. This mixed-method approach provided a robust synthesis of empirical findings and normative policy positions.
A clear international consensus has emerged among leading publishers: generative AI tools cannot be listed as authors or co-authors on scientific papers. AI systems lack legal personality, independent accountability, and the ability to take responsibility for the accuracy, originality, and ethical standards of the work. However, their use as assistive tools is increasingly permitted, provided authors maintain full transparency. Disclosures must typically include the tool’s name and version, specific prompts used, and the extent of AI contribution. These details should appear in the acknowledgments or methods section.
This stance aligns with ethical guidelines from organizations such as the Committee on Publication Ethics (COPE) and the International Committee of Medical Journal Editors (ICMJE). Core principles include: human authors retain full responsibility for content; AI use must be explicitly disclosed; AI cannot receive authorship credit; and strict safeguards must protect manuscript confidentiality.
Generative AI offers substantial advantages for researchers. It significantly enhances linguistic quality, grammar, readability, and manuscript structure. These improvements are particularly valuable for non-native English speakers, helping them overcome language barriers and meet the standards of international journals. Beyond editing, GenAI boosts productivity by automating routine tasks. It can generate initial drafts of sections (e.g., introductions or methodologies), format references according to styles like APA, summarize extensive literature, and identify relevant sources. The tools also support interdisciplinary collaboration by bridging knowledge gaps, generating ideas through pattern recognition, and suggesting hypotheses or experimental designs. In systematic reviews, advanced models can screen articles, extract data, and prepare preliminary syntheses, thereby accelerating the overall research process.
Despite these benefits, serious risks accompany GenAI integration. Plagiarism remains a major concern due to models being trained on vast existing texts. “Hallucination” is another critical issue. Error rates vary by model and task complexity. Algorithmic biases in training data related to culture, language, geography, gender, and institutional affiliation can perpetuate global inequalities. Models often over-cite Western, English-language publications. Over-reliance on AI may diminish researchers’ critical thinking, analytical skills, and original voice, leading to more homogenized academic output. Additional risks include data confidentiality breaches, intellectual property violations, and difficulties in distinguishing AI-generated from human-authored content, which can undermine public and scientific trust.
In peer review, GenAI shows both promise and limitations. It can reduce workloads for editors and reviewers by handling initial screening, plagiarism detection, reviewer matching based on expertise, and preliminary quality assessments. These capabilities may shorten review cycles and reduce certain human biases through consistent criteria application. However, current systems show lower accuracy in interdisciplinary and social science fields. They struggle with nuanced domain-specific judgment and lack the deep contextual understanding needed for evaluating originality, theoretical contributions, and methodological rigor. Ethical concerns are significant: uploading unpublished manuscripts to external AI platforms risks breaching confidentiality and intellectual property rights. Consequently, most major publishers prohibit reviewers from using general-purpose GenAI tools for evaluating or improving review reports.
Considering global developments and Iran’s specific context, the study proposes comprehensive policy recommendations for Iranian journals and national institutions. Iranian publishers should develop clear, transparent, and context-sensitive guidelines that ban AI authorship while allowing limited assistive use with detailed mandatory disclosure. The use of AI for generating core scientific content, figures, tables, or data visualizations should be prohibited or require rigorous justification and review. The Commission for Scientific Journals at the Ministry of Science, Research and Technology should lead by creating unified national policies. Recommended actions include designing mandatory training programs for authors, reviewers, and editors on responsible AI use; investing in indigenous Persian-language large language models trained on Iranian scholarly data; establishing robust data governance frameworks; conducting regular algorithmic bias audits; and implementing continuous monitoring of AI-assisted processes. Domestic models would reduce dependence on foreign systems, address cultural and linguistic biases, and strengthen data sovereignty. Strict protocols for manuscript confidentiality and informed consent are also essential.
The future of scholarly publishing in Iran depends on a balanced, thoughtful integration of generative AI. This approach must harness benefits in efficiency, accessibility, and productivity while protecting core values of scientific integrity, accountability, critical human judgment, and originality. By adopting evidence-based and locally appropriate policies, Iranian journals can mitigate risks, enhance their global standing, and contribute to the responsible advancement of AI in academia.