学术出版中AIGC使用边界指南3.0

1 Background ……………………………………………………………………………………………………………………………… 01 2 Objectives ………………………………………………………………………………………………………………………………… 02 2.1 Prevent academic misconduct and enhance research integrity governance….……………………………… 022.2 Guide relevant stakeholders to reach consensus on the use of AIGC………………………………………..…… 02 3 Principles 3.1 Transparency and accountability………………………………………………………………………….…………………… 023.2 Privacy and data security…………………………………………………………………………..……………………………… 023.3 Fairness………………………………………………………………………………………………………………….………………… 033.4 Sustainable development…………………………………………………………………………………………...……………… 03 6 Conclusion 7 Acknowledgements..………...............................................................................................................…………………… 11 Background In recent years, arti�icial intelligence (AI) technologyhas been developing rapidly, especially with the releaseof ChatGPT, the AI chatbot in November 2022. Arti�icialIntelligence Generated Content (AIGC) has entered thepublic eye and is widely used. It is clear that AI is gainingtheability to generate�luent language,making itincreasingly dif�icult to distinguish the mass of generat-Therefore, it is crucial to develop guidelines that clearlyde�ine the boundaries of AIGC usage in the academiccommunity.Currently, various national or regional policymakers,publishers,and other relevant organizations(theCommittee on Publication Ethics (COPE), the Interna- The main concern of the research community is thatscientists, researchers, and students may fraudulentlypresent AI-generated text as their own or simply useAIGC to produce unreliable research results. LargeLanguage Models (LLMs) work by learning statisticallanguage patterns from large online text databases.CONSORT-AI for clinical trials; SPIRIT-AI for clinicaltrial protocols; TRIPOD+AI for predictive models, etc.Therefore, based on an extensive review and study ofexisting research and exploration in the industry, we arecommitted to establishing a framework and guidelinesthat outlines the fundamental principles of best practice stored, and used by the model provider for subsequent3.4 Sustainable development training, raising potential privacy and data security considerations. To mitigate these risks, we recom-mend adhering to data minimization principles andapplying appropriate de-identi�ication measures. It isalso advisable to avoid submitting sensitive informa-Themultidisciplinary nature of AI demonstratesimmense potential in addressing global challenges suchas the United Nations Sustainable Development Goalsand carbon neutrality. However, while empoweringsocietal progress, the substantial energy consumption practically valuable knowledge, ensuring data quality at 3.3 Fairness The utilization of AIGC should be under the principle offairness to avoid bias. As AI has the risk of replicatingand amplifying bias, potential sources of bias should becarefully assessed and reviewed in the process of train-ing data selection, algorithm design, model generation,model training—thereby reducing unnecessary energyconsumption.Sustainable development should be a core principle ofAIGC itself. To avoid resource waste, excessive reliance mustremain a key guiding principle,driving thesustainable development of the technology. Behavioral framework/practice guideline AIGC tools can provide assistance (services) at variousstages of research and academic publishing. In order toon compliant and responsible use of AIGC. foster a conducive research environment, to addresspotential issues, and to prevent/reduce misuse of AIGC,4.1 Research and writingThis section mainly provides guidance to researchers on ensure that researchers use only valid, unbiased materi-al and prevent the dissemination of false, biased, ordiscriminatory information.lse, biased, or discriminato- 4.1.2 Statistical analysis AIGC tools to interpret data, calculate statistical indica-tors, perform simple data analysis, and describe statisti-cal results. However, AIGC cannot replace the research- and sentiment tendencies of the public or experts on covers types of images including video and animation 4.1.3 Charting Charting and Image Generation: Based on the character-istics of the data and the intended purpose, AIGC toolscan recommend the most suitable type of statisticalgraph for the application scenario. This helps presentstatistical results in a clear and effective manner, allow-ing researchers to convey their �indings more ef�icient-ly. As a result, AIGC saves time in graph creation andenhances overall writing productivity. However, all(such as video stills), photography, scienti�ic diagrams,photo illustrations and other collages, as well as editori-al illustrations such as drawings, cartoons, or other 2Dor 3D visual representations. It is not acceptable toenhance, obscure, move, remove, or introduce a speci�icfeature within an image. Adjustments of brightness,contrast, or color balance are acceptable only if they are