A new friend? Remember to click on Chinese Ethics to follow me online~

Here, be with ethics enthusiasts~

Legislative ideas for the ethical review system of artificial intelligence

*Author Li Xueyao

Professor of Kaiyuan Law School of Shanghai Jiaotong University

Planning Committee Member of the Institute of Chinese Law and Society, Shanghai Jiaotong University

Artificial intelligence ethics have a symmetry and continuity relationship with biomedical ethics, but the responsibility subject, review object, and problem awareness and weight of substantive principles and rules (such as the principle of autonomy) are different from that of biomedical. ethics. The biomedical ethics review process is reflected in a compromise between "mass moral intuition" and "unbanned technological innovation". Artificial intelligence ethics review still has this nature, but its "mass moral intuition" can be directly implemented through technical channels. The ethical challenges of artificial intelligence are mainly focused on the deployment and application stages, while biomedical ethics mainly focus on the R&D and development stages. From the above analysis, the artificial intelligence ethical review system should be legislated separately within the framework of science and technology ethical governance, and adopt differentiated biological aspects in terms of the responsible subject of ethical review, the conditions for initiating review, expert composition, rules of procedure and the legal effect of the review conclusion. Legislative ideas for medical ethics.

Seeking enlightenment from the biomedical ethics system

Article 7 of the "Regulations on the Recommendation and Management of Internet Information Service Algorithms" issued by the Cyberspace Administration of China in 2022 is the beginning of building an artificial intelligence ethical review system in the legislative sense of my country. In July 2023, the State Cyberspace Administration of China and seven other ministries and commissions promulgated the "Interim Measures for the Management of Generative Artificial Intelligence Services" (hereinafter referred to as the "Interim Measures"), which exemplifies relevant principles of artificial intelligence ethics in the sense of substantive law rules. Subsequently, the Ministry of Science and Technology and other units also promulgated the "Technology Ethics Review Measures (Trial)" (hereinafter referred to as the "Ethics Review Measures") with the nature of a normative document and listed it in the appendix "List of Science and Technology Activities that Need to Carry out Ethics Review and Review" Several situations where artificial intelligence ethics review is necessary.

Nowadays, for regulatory departments and compliance entities at all levels, how to transform artificial intelligence ethics from a framework of ethical principles into actionable, predictable and calculating ethical compliance practices is a practical issue that needs to be solved urgently. Recently, relevant central departments are soliciting opinions from all walks of life on the issue of formulating special regulations on artificial intelligence ethics review. So, how should my country build an artificial intelligence ethical review system under the idea of ​​"operable, computable and predictable"? Given that biomedical ethics and its ethical review practice are relatively mature, it can be used as an important reference object and conduct relevant discussions. Generally speaking, we can discuss it from four aspects: First, what are the commonalities between the two. Second, what are the differences between the two and what are the characteristics of artificial intelligence ethics. Third, what lessons should be learned from the institutional practice and theoretical discussion of biomedical ethics. Fourth, what aspects of the institutional tradition of biomedical ethics can be directly inherited and learned by the practice of artificial intelligence ethics.

The continuity of artificial intelligence ethics and biomedical ethics

What has always been controversial in the applied ethics community is whether the development of artificial intelligence technology raises new and unique ethical questions, or does artificial intelligence ethics just repeat the same ethical dilemma encountered in more mature fields, such as the field of biomedical? In this regard, most scholars who have studied biomedical ethics will argue that there is a high degree of consistency or continuity between the two, and believe that it should be integrated into the theoretical system of applied ethics.

If we compare the four principles of biomedical ethics, dignity, benefit, harmlessness and justice, with the legislative and policy documents of various countries and the principles of the ethical entity of artificial intelligence listed in the academic community, we can indeed draw the following conclusion: There is a strong degree of symmetry and consistency between the basic principles of intelligent ethics and biomedical ethics. (See Table 1) In other words, within the legal framework, analogy and other methods can be used to deduce the relevant principles and rules of the artificial intelligence ethical review system from the substantive rules of biomedical ethical review.

Differences between artificial intelligence ethics and biomedical ethics

The above article reflects on the idea of ​​overemphasizing the uniqueness of artificial intelligence ethics. However, in the practice of ethical review, we also need to be vigilant about the conservative idea of ​​completely treating biomedical ethics and artificial intelligence ethics as one, and simply replicating the practice of biomedical ethics review in the practice of artificial intelligence ethics compliance. The author once used the conceptual tool of "the essence of things" to initially summarize the three constraints that distinguish artificial intelligence ethics from biomedical ethics due to its technical characteristics: the technical embeddability of moral rules, stronger scenarios, and dependence on Procedural nature of the technical process. Here we continue the discussion along the theoretical extension line in this article:

First, the differences in the unity and "generalization" of ethical responsible parties in compliance entities. In addition to developers and providers, users of artificial intelligence will become the focus of ethical challenges; the scope of the latter can be understood as all natural and non-natural person subjects in the world. On the one hand, the use of artificial intelligence ethics by different subjects can exponentially improve the operation efficiency of the entire society, but on the other hand, the ethical controversy it causes is complex, and it is unimaginable to exclude it from the scope of ethical review. In other words, compared with biomedical ethics, artificial intelligence is suitable for almost any human activity, which provides countless possibilities for its use, which determines that it cannot be simply reduced to the occupation of scientific researchers or doctors like biomedical ethics. ethics. It will challenge the ethical and moral rules that have been formed in all fields of human beings, and all users have the potential to become potential ethical challengers.

Of course, this will lead to two discussions: one is the ethical conflict in the use stage of artificial intelligence, which may ultimately need or can be achieved through legislative procedures or judicial review, but considering the complexity, uncertainty and efficiency improvement of specific application scenarios. The obligations of ethical review can be authorized to platform-type and public management users through various forms of legal authorization, and most of the obligations to solve ethical difficulties can be allocated to the ethical review committee established by self-regulatory subjects in accordance with the law. Second, it is necessary to distinguish between different types of moral dilemma scenarios caused by technological innovation. The ethical debate in the movie "I Am Not a God of Medicine" is caused by whether generic drugs can be used and how they are used, but the root cause of their dilemma is the value conflict between administrative order maintenance and curing diseases and saving lives, rather than how this technology affects individuals and humans. The issue of autonomy.

Second, the differences in the principles and rules focus on compliance content. Taking the principle of autonomy as an example, there is no doubt that the autonomy of human collective and individuals is the key to the ethical principle of artificial intelligence. First, the autonomy in the sense of individual freedom, including the protection of privacy and property rights, etc.; second, the autonomy in the overall sense of human beings, including highly autonomous artificial intelligence systems, should be designed to ensure that their goals and behaviors are in the entire operation process. Keep in line with human values. Although traditional biomedical technology research and development and application, such as deadly biological weapons, may also cause the danger of human beings being destroyed as a whole or being controlled by other non-humans, artificial intelligence systems may take over and threaten the ethical challenges reflected in the entire human autonomy. , not the focus of traditional biomedical ethics.

Take the principle of justice as an example. Since artificial intelligence systems have deeply intervened and gradually replaced human decision-making, they are essentially different from biomedical technology itself only the decision-making object and the interest distribution object. Therefore, the research and development of biomedical technology involves distribution justice at most, and is generally what is considered in the redistribution stage, and is not the object of thinking in the research and development stage. But the operation of the artificial intelligence system itself will dissolve or alienate the existing legal procedures and systems. Compared with all research objects handled in the field of applied ethics in the past, the application of artificial intelligence and its transformative impact on society are more extensive and far-reaching. Therefore, the research, development, deployment and use of every cutting-edge artificial intelligence technology is necessary to carry out self-censorship from the perspective of human justice or whether the existing social order of mankind can be continued.

Third, the distinction between "immorality" and "morality" in compliance objects. The biggest difference between the review objects of artificial intelligence ethics and biomedical ethics is the requirements of "immorality" and "ethicality". Recently, the substantive content of the international principle of artificial intelligence has gradually been condensed and elaborated as the "trustworthy principle", and "trustworthy" can actually be translated as the expression of "morality".

In the field of biomedical science, biomedical technologies and products in the sense of non-artificial intelligence technology have distinct "non-ethical" characteristics, so the "technological neutrality principle" can be applied smoothly in legal compliance. The difference between artificial intelligence and traditional biomedical research and development and application is that its products and services are not just hardware devices, but are usually programs and applications in extremely complex systems, such as health and welfare data generation systems and tax systems. Therefore, it is obviously difficult to simply apply the principle of technology neutrality for artificial intelligence technologies and products, especially generative artificial intelligence systems.

Key points of building an artificial intelligence ethical review system

Through the above analysis, the legislative ideas of my country's artificial intelligence ethical review system can be developed from the following aspects.

First, legislative model. Through the previous analysis, it is necessary to distinguish the legislation of artificial intelligence ethics from the legislation of biomedical ethics. The artificial intelligence ethical review system should be legislated separately within the framework of science and technology ethics governance, and to make normative requirements that are different from biomedical ethics in terms of the responsible subject of ethical review, the conditions for review initiation, expert composition, legal effect and rules of procedure of review conclusions. .

Specifically, in the current ethical review system, in addition to the "Ethical Review Measures" which are mainly based on biomedical ethical review scenarios, a parallel administrative regulations or normative document specifically targeting artificial intelligence ethics should be drafted as soon as possible. At present, there is a legislative idea that the "Ethical Review Measures" is used as the superior law for the relevant artificial intelligence ethical review system. Since the existing "technology ethics" and its related ethical review systems (such as committee composition) have inherited the "biomedical ethics review" system, in order to prevent the unknowable inertial thinking brought by the latter's review ideas, it is necessary to Draft a normative document or administrative regulations that are not subject to the restrictions of the "Ethical Review Measures" in the short term. Considering that the practicality of the ethical principles of artificial intelligence involves the competitive advantages of China's artificial intelligence technology and industry, and involves a large number of procedural norms and authorized norms, it is necessary to draft a single-line "Artificial Intelligence Ethics" in the short and medium term. When the time is right, it will be compiled as the content of the general "Artificial Intelligence Law".

Second, the specific terms and contents. Given the embeddedness, scenario and procedural characteristics of artificial intelligence ethics, it requires more bottom-up and evolutionary rule generation modes, so in specific ethical judgment scenarios, effective communication between engineers and ethical experts and legal experts. , confrontation, reaching consensus and ultimately transforming into behavior is extremely important. Therefore, in terms of designing relevant clauses for ethical review, it is not advisable to place too many clause design resources on the content design of substantive ethical principles, but should focus on the design of three clauses from a procedural perspective:

First, the main body, responsibilities and responsibilities configuration of the ethical review committee. On the one hand, unlike biomedical technology, the main force in research and development of artificial intelligence technology has gradually transformed into enterprises, rather than universities and scientific research institutions. Therefore, the competent departments of the Ethics Review Committee should change from being led by the competent departments of scientific research institutions and universities to being led by the industry and information departments. On the other hand, from the perspective of reasonable distribution of expert powers and administrative powers, more authoritative clauses are designed. At the same time, we must also cooperate to clarify the gatekeeper responsibilities of "platform development and application enterprises" in the process of artificial intelligence ethical governance, and authorized confirmation of the relevant rules of their self-creation.

The second is about the composition of the ethical review committee. In order to better form an "opposite setting", the committee members should focus on the selection, composition, qualifications and dynamic adjustments of members. Among them, considering the "ethical artificial intelligence" artificial intelligence ethical review goals and technical achievability, during the ethical review process, we should focus on increasing the proportion of artificial intelligence technology experts. Artificial intelligence ethical review still has the characteristics of "popular moral intuition", but it can be directly implemented through technical means, which should be reflected in the qualifications and composition of the ethical review committee.

The third is the rules of procedure regarding the Ethical Review Committee. Pay attention to the design of evaluation procedures and decision-making procedures. Related to the second point, given the technical need for embeddability of artificial intelligence ethics, a discussion process suitable for technicality can also be set up in the evaluation program design to prevent the ethical review procedures from being idle and eventually becoming "popular moral intuition". Problems with restricting the advancement of artificial intelligence technology” tools.

Fourth, the conditions for the initiation of the artificial intelligence ethics review procedure. The focus of artificial intelligence ethical review should not be on the research and development stage, and the responsible subjects should be identified only as developers and providers. The responsibility for ethical review should be allocated more to users, especially decision-making artificial intelligence systems. Industry user or platform user. This change in the focus of ethical review is not only caused by the characteristics of artificial intelligence technology, but also the need to optimize the allocation of responsibilities to better promote the research and development of artificial intelligence technology and its industrial development.

The original article was published in the 10th issue of "Exploration and Controversy" in 2024.

Original quote: Li Xueyao. Legislative ideas of the artificial intelligence ethical review system [J]. Exploration and Controversy, 2024, (10): 35-38 177.

Welcome to recommend

China Ethics Online welcomes all authors to recommend manuscripts to us. The manuscript can be either an article that has been published in a journal, newspaper, website or conference, or an article that is specially created for Chinese Ethics Online.

Depending on the type of manuscript, please note:

1. Academic papers must comply with academic standards, strive for innovation, and have certain academic value.

2. Current commentary articles should focus on the combination of theory and practice, and try to substitute them into the perspective of ethics and related philosophical disciplines.