Xinhua News Agency, Tianjin, August 30th (Reporter Sister Liang) Algorithms are deeply embedded in people’s lives. However, the negative effects such as the killing of big data and the disclosure of personal privacy have caused people to worry about the abuse of algorithms.
On March 1st this year, the Administrative Regulations on Internet Information Service Algorithm Recommendation was formally implemented. Six months after the implementation of the new regulations, what new progress has China made in the field of algorithm governance? The guests and other experts and scholars attending the sub-forum of Algorithm Governance of China Network Civilization Conference in 2022 gave their own answers.
How effective is the algorithm governance for half a year?
As the foundation of digital economy, algorithm has been widely used for many years. Liu Xianglong, deputy director of the State Key Laboratory of Software Development Environment, Beihang University, thinks that the algorithm has changed from manual rules to data-driven. "The current algorithm is deeply coupled with big data. Because of the uncertainty of data and the unclear boundary of algorithm capability, the algorithm behavior will be uncontrollable."
On March 1st, the Regulations on the Administration of Internet Information Service Algorithm Recommendation (hereinafter referred to as the Regulations) jointly issued by the National Network Information Office and other four departments was officially implemented. This is China’s first targeted algorithm recommendation rules and regulations in the field of Internet information services. A number of interviewees said that the promulgation of the "Regulations" filled the gap in the field of Internet governance, and China entered the stage of systematic governance of algorithms.
According to Liu Xiaochun, executive director of the Internet Rule of Law Research Center of China Academy of Social Sciences, the implementation of the Regulations means that China is at the forefront of the world in the exploration of algorithm governance. "What is more commendable is that the Regulations respond to practical problems such as algorithm content governance and minor protection, and it is a governance plan with China characteristics."
At the forum, many guests shared the achievements of the "Regulations" in the past six months: launching the special action of "Clear and Comprehensive Algorithm Management" in 2022, and conducting on-site inspections on some Internet companies; Publish the first batch of domestic Internet information service algorithm names and filing numbers, including many large Internet enterprises and products; It is clearly stated that "if the user chooses to close the algorithm recommendation service, the algorithm recommendation service provider should immediately stop providing related services" and return the choice of algorithm use to the user … …
In addition to the actual governance actions, Liu Xianglong believes that the guiding impact of the implementation of the Regulations is equally important. "This releases a signal of top-down system governance of the algorithm. Internet companies should actively regulate their own behaviors to ensure the rights of users and fair use."
On August 28th, at the National Convention and Exhibition Center (Tianjin), participants visited the exhibition at the 2022 China Cyber Civilization Conference. Xinhua News Agency reporter Sun Fanyue photo
How to solve the "black box" technical problem of algorithm?
At the sub-forum of algorithm governance, Shen Huawei, director of the Research Center of Data Intelligent System of Institute of Computing Technology, Chinese Academy of Sciences, pointed out that there are some misunderstandings about the filing of Internet information service algorithms implemented simultaneously with the Regulations, mainly the content of filing. "The content of the algorithm filing in China includes three parts: the subject information, algorithm information, product and function information of the algorithm, and does not involve the data and codes used by the algorithm."
Many interviewed experts and scholars believe that the opacity of data, decision-making mechanism and design intention makes the algorithm that relies on data for deep learning and analysis a "black box". Take taking a taxi as an example, every passenger hopes to get a taxi as soon as possible. However, the single dispatching algorithm of the platform may consider how to allocate vehicles to achieve the optimal global capacity, and then design subsidies to maximize profits. There is a certain deviation between the optimization goal of the algorithm and the user’s demands.
Therefore, algorithm recommendation service providers, as an important part of China’s algorithm governance system, are of great significance to the supervision of algorithms before and during the event.
Shen Huawei introduced that algorithm recommendation service providers should conduct self-assessment of algorithm security before online application of the algorithm to eliminate possible risks. At present, there is still a long way to go to effectively supervise the "black box" algorithm before the technical development and supervision tools are mature, and the algorithm security assessment can be an important supplement to the algorithm governance.
On August 28th, at the National Convention and Exhibition Center (Tianjin), participants took photos of the exhibition at the 2022 China Cyber Civilization Conference. Xinhua News Agency reporter Li Ran photo
How do algorithms and people get along harmoniously?
At the forum, many experts said that the current guidelines and regulations on algorithm governance issued by China are actually aimed at the "code of conduct" of algorithm use rather than the technical bottom.
With the frequent occurrence of problem algorithms, should people be in a passive position in the process of getting along with algorithms? In the practice of algorithmic governance in China, can users find the initiative?
Liu Xiaochun pointed out that the Regulations clearly require that "if a user chooses to close the algorithm recommendation service, the algorithm recommendation service provider shall immediately stop providing relevant services". "This has improved the active position of users in algorithm services, which means that users can have the right not to be selected, guided or even entangled by algorithms."
Liu Xianglong believes that people and algorithms are essentially a relationship between "entrustment" and "agency" — — Users give personal data to the algorithm and entrust their agents to make the best decisions themselves. But in fact, due to the technical threshold of the algorithm itself, users lack the corresponding technical identification ability. "This requires as ‘ Middleman ’ The algorithm service provider compacts the main responsibility, practices social values, makes the algorithm data more transparent and open, and gives users a sense of trust. "
In addition, under the background of systematic governance of algorithms, users, as important participants, should also rationally recognize algorithms, neither fear technology nor relax the reasonable requirements for algorithms, and form a joint force with the regulatory authorities to promote the transition from "algorithm selection" to "algorithm selection".
