GB/T 46799-2025Artificial intelligence social experiments - Evaluation guidelines (English PDF)

Foreword

This document complies with the provisions of GB/T 1.1-2020 "Standardization Work Guidelines Part

1.Structure and Drafting Rules of Standardization Documents". Drafting. GB/T 46801 "Guidelines for the Design of Social Experiments in Artificial Intelligence" and GB/T 46802 "Guidelines for the Implementation of Social Experiments in Artificial Intelligence" and GB/T 46799, "Evaluation Guidelines for Social Experiments in Artificial Intelligence," together constitute the national standard system supporting the conduct of social experiments in artificial intelligence. Please note that some content in this document may involve patents. The issuing organization of this document assumes no responsibility for identifying patents. This document was proposed and is under the jurisdiction of the National Working Group on Basic Standardization of Social Application and Evaluation of Intelligent Technologies (SAC/SWG35). This document was drafted by: China National Institute of Standardization, Tsinghua University, Zhejiang University, Runze Intelligent Computing Technology Group Co., Ltd., and China Mobile. Information System Integration Co., Ltd., Hefei Qinghe Intelligent Technology Co., Ltd., Renmin University of China, Shandong Provincial Computing Center (National Supercomputing Center) South Center), Zhejiang Sci-Tech University, Xi'an Jiaotong University, North China University of Technology, Weikun Intelligent Technology (Shanghai) Co., Ltd., Guizhou University, Public Security Ministry of Public Security Security and Police Electronic Products Quality Testing Center (First Research Institute of the Ministry of Public Security), China Electronics Technology Group Corporation Electronic Science Research Institute Research Institute. The main drafters of this document are. Su Jun, Zhao Wenhui, Cao Lili, Zhou Chaonan, Ru Peng, Huang Cui, Wang Yun, Li Li, Zhou Biyu, Xie Qijun, and Xu Jingting. Huang Keman, Sun Jinyang, Gao Zongbang, Zhang Pengfei, Shen Chao, Wang Xu, Wang Guiling, Chen Jing, Zhang Xian, Zhu Jing, Shen Cheng, Xia Fei, Wu Yueyue, Zhang Wenhong, Wang Yingxue Gao Yunfeng, Wang Yu, Yang Heng.

Human society is moving towards an intelligent society supported by artificial intelligence as a key technology. Summarizing and forming experiences and norms for the development and governance of an intelligent society is crucial. Laws and theories, proactively exploring standards and norms for the development and governance of an intelligent society, and improving the systems and mechanisms adapted to the development and governance of an intelligent society will have It effectively supports the modernization of the national governance system and governance capabilities. Artificial intelligence social experiments are a key method for exploring the development and governance paths of intelligent societies, and also a standardization approach for the development and governance of intelligent societies. One of the fundamental methods. Through conducting long-term, wide-area, multidisciplinary artificial intelligence social experiments, it systematically identifies, defines, observes, and evaluates... This paper analyzes the impact of specific intelligent technologies on social application scenarios, examines potential governance challenges, summarizes countermeasures, and formulates solutions and norms. It is of great significance to reasonably regulate the social risks of technology application and promote the healthy development of technology. my country has established a national standards system to support social experiments in artificial intelligence. Within this system, GB/T 46799 "Artificial Intelligence..." The "Guidelines for Evaluating Artificial Intelligence Social Experiments" aims to provide general principles, evaluation index systems, and evaluation processes for evaluating artificial intelligence social experiments. The suggestions and information provided by the process offer reference and guidance to relevant technical, application, and research entities engaged in social experiments related to artificial intelligence. The development of experiments related to artificial intelligence society will improve the scientific rigor and standardization of experimental evaluation. Guidelines for Evaluating Social Experiments in Artificial Intelligence

1 Scope

GB/T 46799-2025 is the Chinese national standard covering evaluating what an AI social experiment found - the indicators of effect on employment, behaviour, fairness and wellbeing, the attribution of a change to the technology rather than to everything else happening at the same time, and the reporting. First edition, with the design guidelines GB/T 46801-2025 and the implementation guidelines GB/T 46802-2025. It was issued on 2 December 2025 and has been in force since 2 December 2025, as a first edition. The document is under the responsibility of the Standardization Administration of China. This page is published from the official record of the 2025 edition; the clause text of a standard this recent is not yet in circulation, and the figures, limits and tables it contains are those of the document itself, delivered in full with the English translation.

This document establishes the general principles for evaluating social experiments in artificial intelligence, and provides an evaluation index system, value selection rules, evaluation process, and evaluation guidelines. Guidance on pricing results. This document is intended to guide stakeholders, including those involved in the application, research, and technology of AI social experiments, in evaluating AI-driven social experiments. The experiment also provides practical reference for relevant entities to conduct social impact assessments on the application of intelligent technologies.

2 Normative references

The contents of the following documents, through normative references within the text, constitute essential provisions of this document. Dated citations are not included. For references to documents, only the version corresponding to that date applies to this document; for undated references, the latest version (including all amendments) applies. This document.

GB/T 46802-2025 Guidelines for the Implementation of Social Experiments in Artificial Intelligence

3 Terms and Definitions

The terms and definitions defined in GB/T 46802-2025 apply to this document.

4 General Principles

4.1 Scientific nature The evaluators, based on the relevant requirements for the design and implementation of social experiments in artificial intelligence, considered experimental methods, scenario adaptation, experimental intervention, and ethical review. Based on the characteristics of the experiment, scientifically determine the evaluation indicators, accurately determine the value rules, and integrate qualitative evaluation with quantitative calculation, static analysis with dynamic analysis. The evaluation results are comprehensively presented by combining qualitative and quantitative analysis.

4.2 Accuracy The evaluated party shall truthfully provide all documents required for the evaluation, ensuring the authenticity and accuracy of the documents. The evaluator shall conduct the evaluation objectively and accurately. The evaluation involves objectively reviewing and analyzing all documents and related materials, and reporting the evaluation findings and results truthfully and accurately. The evaluation process is documented in a complete form. The evaluator archives the documents supporting the evaluation results to ensure that the relevant results are traceable.

4.3 Comprehensiveness The evaluation should analyze the three procedural stages of application, scientific measurement, and comprehensive feedback (see section

4.2 of GB/T 46802-2025). In addition to considering the sex and effectiveness, it is also necessary to take into account the basic principles of organizing and implementing social experiments on artificial intelligence, and to pay full attention to the impact of intelligent technology on micro-innovations during the application process. The impact on individuals, meso-level organizations, and macro-level social systems.

4.4 Sociality The evaluation should fully consider whether the experiment has formed empirical patterns and policy systems for the development and governance of an intelligent society, and whether it can effectively support... Support the development and governance of a smart society.