Algorithm recommendation service has been "tightened", and traffic fraud and hot search control have been cured.

Algorithm recommendation is a "double-edged sword"

The report "Opportunities and Challenges of Digital Content Governance in the Age of Artificial Intelligence" released by Tencent Security Strategy Research Center and Cyber Research Institute shows that algorithm accurate recommendation has occupied the dominant position in information flow distribution, and the algorithm can realize digital content aggregation and accurate recommendation, and quickly match information and people. Personalized content push based on algorithm not only provides users with accurate and personalized services, but also brings problems such as increased national security risk factors, increased risk of bad information flooding and increased difficulty in protecting traditional rights. It has become a "double-edged sword".

Cui Lili, a researcher at Shanghai University of Finance and Economics, said that the Internet platform, as an advanced technical representative to eliminate information asymmetry, has the responsibility to maintain the objectivity of technical means, reflect the natural state of traffic and present real public opinions. We should not cheat on traffic because of commercial interests. In the past, some news and information related to commercial interests have been manipulated, and the commercial interests obtained in this way are unfair or even illegal in the market.

Put forward new requirements for algorithm "kindness"

According to the draft for comments, the management regulations will apply to companies that provide Internet information services by applying algorithm recommendation technology in China. This also means that the algorithm recommendation technologies of mainstream Internet companies and platforms, such as short video platforms with content recommendation algorithms as their core competitiveness, UGC (User Produced Content) platforms, e-commerce and social platforms that recommend content or advertisements, are all included here.

Cui Lili said that whether it is UGC or PGC (professional content production) platform, there are actually corresponding rules for content review and audience recommendation. Internet information platform has replaced the traditional media in the past in form, so it may have certain public facilities, so the generation, review and distribution of information should meet the requirements of fairness, objectivity and truth.

the exposure draft points out that algorithm recommendation service providers should adhere to the mainstream value orientation, optimize the algorithm recommendation service mechanism, actively spread positive energy, and promote the "upward goodness" of algorithm application. The algorithm recommendation service provider shall regularly review, evaluate and verify the mechanism, model, data and application results of the algorithm, and shall not set up an algorithm model that induces users to indulge or spend too much, which violates public order and good customs.

Wang Juan said that the draft for comment put forward the requirements for the algorithm to be "good" in terms of subject value orientation, personal data use and industry self-discipline, put forward the filing responsibility of the algorithm provider, clarified the main responsibility of the algorithm recommendation service provider as the implementation of algorithm security, and also put forward the retention requirements for the information such as the algorithm recommendation service log, which should be kept for no less than six months, and provided when the relevant law enforcement departments inquired according to law, thus strengthening the retrospective verification link.

the draft for comment makes it clear that the algorithm recommendation service provider should implement the main responsibility of algorithm security, establish and improve the management systems such as user registration, information release review, algorithm mechanism review, security assessment and monitoring, emergency handling of security incidents, data security protection and personal information protection, formulate and disclose relevant service rules for algorithm recommendation, and equip professionals and technical support suitable for the scale of algorithm recommendation service.

from the user's point of view, the algorithm also puts forward many requirements for the data interaction mode. According to the exposure draft, the algorithm recommendation service provider should inform users of its algorithm recommendation service in a significant way, and publicize the basic principle, purpose and intention, operation mechanism of the algorithm recommendation service in an appropriate way. At the same time, the algorithm recommendation service provider should provide users with the function of selecting, modifying or deleting user tags for algorithm recommendation services, as well as options that are not specific to their personal characteristics, or provide users with the option of conveniently closing the algorithm recommendation service.

As for the chronic diseases such as "killing big data", the exposure draft emphasizes that if the algorithm recommends service providers to sell goods or provide services to consumers, they should protect the legitimate rights and interests of consumers, and should not use the algorithm to implement unreasonable differential treatment and other illegal acts on trading conditions such as trading prices according to the characteristics of consumers' preferences and trading habits.

Online hot search will be strictly controlled

In recent years, online hot search has become a real-time weather vane and a distribution center for public opinion, and the draft for comments has specially regulated hot search, making it clear that algorithm recommendation service providers should strengthen the ecological management of algorithm recommendation service pages, establish and improve the mechanism of manual intervention and users' independent selection, and actively present information content that conforms to mainstream value orientation in key links such as home page first screen, hot search, selection, list category and pop-up.

Wang Juan said that the draft for comment can be described as a "heavy blow" to hot search and false traffic, and the core is still that the Internet is not a place of extra-legality, establishing a clear cyberspace and establishing good social values, and the operation mode of various "rhythmic" capital superimposed traffic such as Weibo Hot Search List will be greatly challenged.

in the face of external doubts, Weibo recently issued the management rules of hot search, and responded to such questions as "paying for hot search" and "paying for hot search", saying that hot search is calculated according to the real behavior of users in Weibo, and a real-time list is formed according to data indicators such as search volume, blog posting volume and interaction volume. The list algorithm includes strict drainage army and anti-garbage mechanism to ensure fairness and objectivity. At the same time, in the first half of the year, hot search also took "reducing the proportion of entertainment" as the main regulation goal.