"AI as an Auxiliary Tool, Responsibility Lies with Researchers"...Scientific Community Raises Concerns over Government Research Ethics Guide [Reading Science]

Ministry of Science and ICT Finalizes 7 Key Recommendations for AI Research Ethics

Researchers Voice Concerns: "AI Guidelines Diverge from Reality... Security and Administrative Burdens May Undermine Research Autonomy"

Even when artificial intelligence (AI) is used to write papers and analyze research findings, the final judgment and responsibility must remain with the human researcher. When evaluating national research and development (R&D) projects or reviewing academic papers, reviewers must not input unpublished materials into external AI systems. While the government has presented a code of research ethics for the AI era, some in the scientific community point out that it does not sufficiently reflect the realities of today’s research environment. There are concerns that efforts to enhance transparency in AI use could lead to new administrative burdens and overly restrict research autonomy.

A scene in a laboratory where a human scientist and an AI are reviewing a research paper together. AI-generated image.

The Ministry of Science and ICT finalized the “National R&D AI Research Ethics Guide” through a review and resolution by the Operating Committee of the National Science and Technology Advisory Council and released it on the 20th. As AI is introduced throughout all phases of research, raising risks such as inaccurate information generation, bias, and research information leaks, the government has established ethical guidelines that must be followed during national R&D projects.

Disclosing AI Use and Human Verification as Final Step

The guide contains seven key recommendations: fact-checking, independent judgment, ensuring transparency, guaranteeing result reliability, compliance with policies and regulations, ensuring security, and protecting personal information. The central principle is that, even when using AI, researchers must directly verify and judge the results themselves.

For national R&D project evaluations and paper reviews, it is stipulated that evaluators must not input materials under review into external AI systems. The guide also warns against attempts by researchers under evaluation to manipulate AI assessment outcomes through means such as hiding prompts in documents—so-called “hidden prompts.” Additionally, a self-diagnosis checklist for ethical issues and a disclosure form for AI usage have been provided for researchers.

Lee Joon-sang, professor in the Department of Mechanical Engineering at Yonsei University, stated, “Rather than restricting the use of AI, the guide is meaningful in that it clearly sets forth the principle that AI may be used as a research aid, but the judgment and responsibility belong to the researcher.”

"AI Is Already Involved in Hypothesis Generation... Guidelines Must Reflect Research Realities"

The controversy centers on whether the principle of AI as an auxiliary tool in research truly reflects the changes occurring in real-world research settings.

Yeom Hanwoong, director of the Center for Artificial Low Dimensional Electronic Systems at the Institute for Basic Science (IBS) and professor of physics at POSTECH, pointed out, “AI is now used in much broader ways than simply organizing data, including data analysis and extraction of theoretical models.” He argued that the guidelines fail to reflect the reality that AI is being used throughout the research process, not only for paper writing but also for research planning, hypothesis generation, and methodology decisions.

The principle of restricting AI tools for security reasons is also an issue. Lee Minouk, principal researcher at the Special Structure Research Center at the Korea Institute of Science and Technology (KIST), expressed concern about the performance gap between the internal closed-network AIs used within institutions and the latest commercial AI systems. He warned that restricting access to commercial AI could reduce research quality and drive researchers to use AI by unofficial means.

Park Jungbin, professor in the Department of Biomedical Convergence Engineering at Pusan National University, emphasized that the responsibility of the institution to provide a safe research environment is distinct from the pre-emptive control of researchers’ tool choices, advocating for the need to respect research autonomy. Park also pointed out that requiring disclosure of AI usage could create additional administrative burdens.

On the other hand, some argue that enforcement measures must be clarified to enhance the effectiveness of ethical standards. Ku Jaho, professor of Atmospheric Sciences at Yonsei University, proposed compiling a casebook of AI misuse and overuse, as well as discussing disciplinary measures for clear cases of abuse, such as the use of hidden prompts.

Addressing disparities in AI use based on the size and resources of research institutions was also recognized as a challenge. Park Hanwoo, dean of the College of Social Sciences at Yeungnam University, argued for the need to offer secure large language models (LLM) and AI infrastructure for research as public resources to universities and small research institutions outside the capital region.

Dean Park said, “The important question is not ‘Was AI used?’ but rather ‘What did AI do, and what did humans verify?’ He emphasized, “Research ethics should go beyond rules about what must not be done to consider what should be provided to ensure research is conducted safely.”

This content was produced with the assistance of AI translation services.