MSTI Releases National R&D AI Research Ethics Guidelines, Says Final Responsibility Rests With Researchers
TECHWORLD ·
✦ AI Summary
The Ministry of Science and ICT has finalized and announced the 'National R&D AI Research Ethics Guidelines' in response to the growing use of AI in national R&D.
The guidelines limit AI to a research support tool, say final responsibility and rights to deliverables rest with researchers, and do not recognize AI as an author or inventor.
They present 7 recommendations for researchers, including fact-checking, independent judgment, transparency, security, and personal information protection.
The Ministry of Science and ICT has finalized and announced the 'National R&D AI Research Ethics Guidelines,' which set out compliance standards for researchers, in response to the growing use of AI in national research and development (R&D). The ministry finalized the guidelines after deliberation and approval by the Standing Committee of the National Science and Technology Advisory Council.
The guidelines are aimed at researchers who use AI in national R&D. Their purpose is to support responsible use of AI by researchers throughout the entire national R&D process.
AI use in research settings is increasing. While AI use can improve research productivity, its expansion has also raised new ethical issues.
The ministry identified risks such as AI generating inaccurate information, reproducing existing data bias, leaking unpublished research materials, leaking personal information, and citing sources that are not clearly identified. Accordingly, the guidelines define AI as a research support tool.
The guidelines also state that researchers are the final party responsible for deliverables and the party to whom rights to the deliverables belong. The ministry said this is intended to support responsible use of AI by researchers throughout the national R&D process.
As a basic principle, the guidelines limit AI's role to a research support tool. They also state that when AI is used in research, responsibility for the final deliverable rests with the researcher, and rights to the final deliverable also belong to the researcher. In addition, they set out the principles that AI cannot be recognized as an author and AI cannot be recognized as an inventor.
Accordingly, the recommendations for researchers consist of 7 items: fact-checking, independent judgment, transparency, management of result reliability, compliance with policies and regulations, security, and personal information protection. The guidelines recommend checking whether AI-generated content contains errors, directly verifying whether key materials are omitted, and directly verifying the existence and authenticity of sources. They also advise against uncritical use of AI-generated materials and call for review based on logical validity and specialized knowledge, as well as reconstruction from the researcher's perspective.
The guidelines set out strengthening transparency in the AI use process as a direction. If AI is used during the research process, researchers are advised to disclose the model and version used, as well as the purpose and scope of use, when submitting or presenting the final output.
The guidelines also recommend ensuring logical explainability in the process of implementing research results, even when AI is used. This means researchers should directly verify and reconstruct AI outputs, and should also be able to disclose and explain the fact and method of use at the final output stage.
For research requiring security, the guidelines first require the use of environments and tools approved by the research institution to which the researcher belongs, and they state as a principle that research data containing personal information should not be entered into external AI services. However, if research data containing personal information must unavoidably be used with external AI, security measures such as de-identification should be taken first.
They also separately set out standards to apply in project evaluations and paper reviews. Evaluators may not enter materials under review, such as research proposals, result reports, or papers, into external AI, and they must conduct evaluations based on specialized knowledge and their own judgment. For those being evaluated, the guidelines prohibit actions such as evading review standards with hidden prompts and attempting to distort evaluation results.
In addition, self-check tools for field application have been prepared. Researchers can use an AI utilization standards self-assessment checklist before conducting research, and they can use the same checklist again before submitting the final output. A disclosure form for deliverables is also provided, and it requires the AI tool name, model, usage details, and period to be entered.
Hong Soon-jung, director general of the Performance Evaluation and Policy Bureau at the Ministry of Science and ICT, said that AI is a useful tool that can dramatically improve research efficiency, but that the final party responsible for the results of AI use is the researcher. He added that he hopes the guidelines will help establish a culture of responsible AI use in research settings and serve as an opportunity to secure research integrity in national R&D.
Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407261
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.