Algorithm Recommendation Will Meet New Rules: Hot Search Should Not Be Controlled To Affect Public Opinion. Users Can Close Recommendation

On August 27, the national Internet Information Office released the "Internet information service algorithm recommendation management regulations (Draft)" (hereinafter referred to as the "draft") and solicited public opinions.

"Opinion draft" carries out key rules for algorithm recommendation service providers with public opinion attribute or social mobilization ability, from content recommendation requirements to adhering to mainstream value orientation, to not implementing traffic fraud, hot search and other influencing network public opinion, and refers to information and short video platforms such as microblog, today's headlines, Baidu, Kuai Shou, Shuo Yin, etc.

For the protection of platform employment rights and consumers' rights, the draft requires that the dual algorithm of takeout and online car Hailing platforms for workers and consumers should be standardized to ensure the fairness and transparency of labor algorithm, and resist the algorithm's "big data killing" of consumers.

How to balance the platform's protection of algorithm usage and user rights? A number of experts told the 21st century economic reporter that enterprise compliance should pay attention to the multi-level requirements of technology, management and content value proposed in the draft.

Adhere to the mainstream value orientation, algorithm to the good

Ecological governance of network information content is the focus of Internet Governance in China in recent years.

In December 2019, the state network and Information Office announced the regulations on Ecological Governance of network information content, marking the establishment of a complete ecological governance framework of network information content in China. In January this year, the state cyber information office solicited public opinions on the measures for the administration of Internet information services (Draft for comments on the revised draft) to further specify the requirements for information dissemination. The state network information office, the national anti pornography and anti illegal office and other departments have repeatedly rectified the content of the platform.

However, there are still some problems in network ecological governance. Shang Xixue, deputy director of the network Law Research Institute of China University of political science and law, said: on the one hand, pornographic, gambling fraud and negative energy content flooded the cyberspace; On the other hand, under the abnormal concept of "traffic is the king", the platform uses the algorithm model to deliver "stimulating" topics and "hot spots" information to users through hot search, super call, speculation, aggregation and other information push methods, so that the public will feel mentally uncomfortable when they are passively "eating melons".

"Opinion draft" pointed out that the algorithm recommendation service provider should adhere to the mainstream value orientation, optimize the algorithm recommendation service mechanism, actively spread positive energy, and promote the algorithm application to be better.

Article 9 standardizes the information push behavior of the content service platform in terms of means, requires strengthening the management of information content, establishing and improving the characteristic database for identifying illegal and harmful information, and improving the warehousing standards, rules and procedures. If a synthetic information is generated by an algorithm that has not been marked conspicuously, it shall be marked prominently before continuing transmission.

The governance of content itself is necessary, but in the mass of content, benign content is often buried, vulgar and eye-catching content is easier to "brush" out, falling into the dilemma of "bad money driving good money".

Article 11 of the draft emphasizes "establishing and improving the mechanism of manual intervention and user's independent selection", that is, not only relying on Algorithms for content recommendation, but also increasing the process of manual identification and screening, "actively present the information content in line with the mainstream value orientation in key links such as the first screen of the home page, hot search, selection, list category, pop-up window, etc

"It is equivalent to that in the process of consolidating the main responsibility of the platform, in addition to the governance of rumors and other illegal information, it also adds a handle on algorithm decision-making." Zhu Wei, deputy director of the communication law research center of China University of political science and law, said that the algorithm and logic through which the platform uses data is the key to Internet regulation.

With regard to the protection of minors, Article 16 of the draft specially proposes that "by developing models suitable for minors and providing services suitable for minors' characteristics, it is convenient for minors to obtain information content beneficial to their physical and mental health."

On the same day of the release of the draft, the office issued a notice to strengthen the management of "rice circle" disorder, proposing 10 working measures such as "canceling the list of star artists" and "optimizing and adjusting the ranking rules". The Office pointed out that in the ranking of music works and film and television works, the index weight of sign in, praise and comment should be reduced, and the index weight of work orientation and professional evaluation should be increased. It is not allowed to set up relevant functions to induce fans to make lists, so as to guide fans to pay more attention to the quality of cultural products.

It is forbidden to control hot search and influence public opinion on the Internet

The hot search list is the most concentrated embodiment of attention economy and the most typical sample of agenda setting in the new media era.

Several content regulation and sorting platforms in the draft, such as today's headlines of information and Baidu, fast hand and shaking tone of short video, comprehensive micro blog and Zhihu, etc., are featured in hot search of the list.

As for the hot search ranking, traffic fraud and other issues, the regulatory authorities have repeatedly stepped in to tighten the supervision of the platform.

In June 2020, the state cyber information office instructed the Beijing Internet information office to interview the person in charge of sina Weibo, ordered rectification, suspended the updating of microblog hot search list for a week, seriously dealt with the relevant responsible persons and imposed severe administrative penalties of fine, aiming at the problems of micro blog disturbing Online communication order and spreading illegal information in a public opinion event of Jiang.

The draft requires the algorithm recommendation service providers in terms of algorithm rules and publicity. According to Article 12, strategies such as content de duplication and elimination of intervention should be comprehensively applied, and the transparency and interpretability of rules such as retrieval, ranking, selection, push and display should be optimized. Article 14 stipulates that users shall be informed of the algorithm recommendation service in a prominent way, and publicize the basic principle, purpose, intention and operation mechanism of algorithm recommendation service in an appropriate way.

On August 23 this year, Weibo responded to queries such as "spending money to buy hot search" and "spending money to suppress hot search", and announced for the first time the rules and management rules of hot search products. According to Weibo, hot search is calculated according to the real behavior of microblog users, and a real-time list is formed based on data indicators such as search volume, blog volume and interaction volume. In order to ensure fairness and objectivity, strict drainage and anti garbage mechanism are included in the list algorithm.

As for the regulation of the platform in influencing public opinion and social mobilization, Article 13 of the draft of opinions requires that it is not allowed to use algorithms to falsely register accounts, illegally trade accounts, manipulate user accounts, or falsely praise, comment, forward, and web navigation to implement traffic counterfeiting and traffic hijacking; It is not allowed to use algorithms to block information, over recommend, manipulate the list or ranking of search results, control hot search or selection and other interference information presentation, implement self preferential treatment, unfair competition, influence network public opinion or evade supervision.

At the same time, the draft requires users to have the right to choose. Article 15 mentions that the algorithm recommendation service provider shall provide users with the option that is not based on their personal characteristics, or provide users with the option to conveniently close the algorithm recommendation service. If the user chooses to shut down, it shall immediately stop providing relevant services. This is in line with the "personal information protection law" issued on August 20 about the right of personal information and withdrawal of consent.

Online car hailing and takeout platform double algorithm regulated

In recent years, the labor force of new forms of employment, such as takeaway boys and online car Hailing drivers, has been growing. In 2019, the total number of catering takeout workers has exceeded 7 million. Statistics show that as of July 31, 2021, 241 platforms in China have obtained online car Hailing business licenses, and 3.51 million online car Hailing driver licenses have been issued.

The draft also regulates the algorithms of scheduling decision-making platforms such as online car hailing and takeout, which are constrained by the dual algorithm of workers and consumers.

For workers, the draft requires fair and transparent algorithm. Article 17 proposes that if the algorithm recommendation service provider provides work scheduling services to workers, it shall establish and improve relevant algorithms such as platform order distribution, remuneration composition and payment, working hours, rewards and punishments, and fulfill the obligations of protecting the rights and interests of workers.

In May this year, eight departments including the Ministry of communications and the State Administration of market supervision jointly interviewed 10 platforms, including didi travel, Cao Cao travel, Gaode, and huolala. According to the interview requirements, the platform should actively publicize the pricing mechanism and pricing rules, and continuously optimize the single dispatch system. Caocao, didi and other platforms have proposed to promote the opening of the extraction. Cao Cao said that the current share ratio has been made public through the driver's display and other ways. Didi announced in May that it would promote the drawing of open orders to promote the transparency of drivers' bills.

In July, eight departments including the Ministry of human resources and social security and seven departments including the General Administration of Market Supervision issued the guidance on safeguarding the labor security rights and interests of workers in new employment forms and the guiding opinions on implementing the responsibilities of online catering platforms and earnestly safeguarding the rights and interests of takeaway delivery workers, urging enterprises to formulate reasonable order allocation, piece price, percentage of extraction, remuneration composition and payment, and Working hours, rewards and penalties and other system rules and platform algorithms are proposed to replace the assessment requirements of "the most stringent algorithm" by "algorithm selection".

For consumers, the algorithm needs to limit "big data killing"《 Article 18 of the draft provides that the algorithm recommendation service provider who sells goods or provides services to consumers shall protect the legitimate rights and interests of consumers, and shall not use the algorithm to carry out unreasonable differential treatment and other illegal acts in terms of transaction price and other transaction conditions according to consumers' preferences and trading habits.

Shang Xixue, deputy director of the network Law Research Institute of China University of political science and law, points out that automated decision-making algorithms are becoming more and more frequent in current business application scenarios, causing social risks that threaten the legitimate rights and interests of users.

There are clear provisions in many laws and regulations, such as "big data killing", "online transaction supervision and management measures", "e-commerce law", "anti-monopoly guide of the Anti-monopoly Commission of the State Council on the field of platform economy", and "personal information protection law", which will be implemented on November 1, this year, and other laws and regulations, which require that "differential treatment" should not be applied to individuals in transactions.

Balance algorithm usage and user rights protection

Behind the algorithm rules are commercial interests. Facing the requirements of algorithm recommendation technology, how should enterprises balance the use of algorithms and protect the rights of users?

Wu Shenkuo, executive director of Beijing Normal University International Center for network rule of law, suggested that enterprises should pay attention to the multi-level requirements of technology, management and content value in the draft.

Specifically, it includes three aspects: at the technical level, attention should be paid to the risks that algorithms may bring, especially the tendency of discrimination, falsity and content inconsistent with positive energy; In the aspect of organization and management, it is necessary to build a whole life cycle organization management system from user registration, management to risk assessment, and comprehensively implement the governance from system to person; In the application process of the algorithm, attention should be paid to the value and ethics requirements of the content itself in line with socialist values and the requirements of positive energy transmission.

"From the perspective of user rights protection in a narrow sense, data security and personal information protection need to achieve cross matching between systems under the existing top-level design of laws, so as to realize the comprehensive protection of users' independent rights and other security interests." Wu said.

Automatic decision-making is the most important part of algorithm supervision“ When the platform provides search results of goods or services to users according to their interests, consumption habits and other characteristics, it shall also provide options that are not specific to their personal characteristics. " On the one hand, on the premise of complying with the law, enterprises should design algorithm models and formulate automatic decision-making rules in accordance with the principles of openness, fairness and fairness, and should not provide discriminatory differential treatment to users; On the other hand, under the premise of fulfilling the obligation of "inform consent", the enterprise should guarantee the user's right of consent, rule interpretation and choice for algorithm design and automation decision-making.

With the help of external forces, enterprises may be able to solve problems from a more objective perspective. You Yunting, senior partner of Shanghai Dabang law firm, believes that enterprises can set up a third-party algorithm supervision committee, such as third-party institutions, user representatives, industry association personnel, etc., to regularly carry out the implementation of algorithm ethics within the enterprise, assist the internal self-examination and self correction and audit, and make pertinent suggestions from outsiders, It can effectively reduce the algorithm risk.