Experts and scholars discuss algorithm supervision, how to balance the platform's commercial value and social responsibility as the key to governance

Southern Finance All Media Reporter Wu Li Intern Zhang Ling Beijing reported

The compliance governance of algorithms is becoming an unavoidable problem in the current Internet supervision. As a platform basic mechanism with a wide range of applications, how should the supervision and governance of the algorithm be implemented?

On the afternoon of September 6, the 2022 (8th) China Internet Rule of Law Conference, hosted by the China Internet Association and co -organized by the China University of Political Science and Law, and the Chinese University of Political Science and Law Research Institute, was held in Beijing. Essence At the meeting, many experts discussed around the above issues.

In March 2022, the "Internet Information Service Algorithm Recommendation Management Regulations" jointly issued by the National Internet Information Office, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration of Market Supervision and Administration, provided important rule of law for the comprehensive governance of the Internet information service algorithm Assure. The implementation of the "Regulations" reflects the current development stage of my country's network space governance.

Wu Shenfang, assistant to the dean of the Internet Development Research Institute of Beijing Normal University and doctoral supervisor, introduced the trend of the current algorithm supervision of my country. He pointed out that algorithm supervision will be a process of continuous dynamics and evolution, and it will consolidate the public cognition and common value demands of the public in a specific time.

Wu Shenjuo believes that China's algorithm supervision positioning is particularly special, and emphasizes the main responsibility of related units related to the use of recommendation algorithms. At present, "card -point thinking" is presented during the governance of the algorithm. It requires the focus of the scene recognition, data identification of the algorithm, the establishment of the security mechanism, the strengthening of the manual review mechanism, and the adaptation of ethical design.

In the algorithm supervision of dynamic development, is there a core direction from one to the end? Wu Shenjuo believes that this core pointing should be the transparency of the algorithm based on the system foundation of algorithm classification. Specifically, three demands need to be achieved: "First, the user's algorithm knows the right to know; the right to select the algorithm of the second user;

Whether it is a "governance algorithm" or "algorithm governance", it is essentially the relationship between the relationship between people and algorithms, and it is also a key issue for the communication between social disciplines and technical disciplines. Liu Zhenyu, an associate professor of the School of Law Information Management of China University of Political Science and Law, took the technical mechanism of "credible federal learning" as an example. It introduced the credible federal learning of "trustworthy, credible, effective, controllable, algorithm explanations, and models."

Liu Zhenyu said that the trusted federal learning can help solve the inherent target conflict in the digital economy. At present, it should develop cross -integration of law and multidisciplinary cross -integration. Combining the rigid requirements of the law and local optimization of technology, the federal learning malicious attack behavior is jointly regulated.

How can the algorithm, especially the supervision of algorithm recommendation mechanisms to the practice of social governance? Xiong Wencong, an associate professor at the School of Law School of the Central University of China, combined with the platform algorithm recommendation context.

He takes algorithms to judge copyright infringement as an example, pointing out that copyright infringement is judged by the value that cannot be recognized by technical identification. If the algorithm lacks the source library of comparative works, then it may not be accurately judged. The principle of prevention costs does not blame the ISP to eliminate all infringement content on the platform to avoid non -notifications and malicious lawsuits.

It is worth noting that in judicial practice, the situation in the application of algorithms in civil disputes involved in the network is becoming more and more common, and related judicial governance has accumulated some experience. Sun Mingxi, the president of the three courts of the Beijing Internet Court, pointed out that justice is a model of risk prevention in the pre -regulatory risk prevention of justice. In the judicial link, we must adhere to the principles of technical neutrality and clarify what functions of the algorithm content.

"For example, information recognition may not be the same as information management functions. Although the greater the rights of rights in the broad sense, the greater the responsibility of the rights, but it depends on what rights corresponding to the technology. What kind of responsibilities should be? . "Sun Mingxi said.

- END -