Image source @ Visual China
Text | Moment Business, author | Kuribuki, Editor | Zhou Ye
Algorithmic recommendations are about to usher in an unprecedented revolution.
Recently, the State Internet Information Office, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the State Administration for Market Regulation jointly issued the Provisions on the Recommendation and Administration of Internet Information Service Algorithms (hereinafter referred to as the "New Rules on Algorithms"), which will come into effect on March 1, 2022.
The official promulgation of the new algorithm regulations means that the chaos of algorithm recommendation-related industries will officially bid farewell to the era of barbaric growth, and a more refined regulatory system will carry out long-term governance for algorithm recommendation chaos.
In the past decade, algorithmic recommendation has become the biggest outlet, and many Internet companies have relied on algorithmic recommendation to gain a firm foothold in the market.
Today's algorithm recommendation has become the standard of the digital age, applied to all fields of the Internet, whether it is user portraits, personalized recommendations and other Internet companies to obtain customers, active users, means to enhance user stickiness, or users online shopping, ordering takeaways, watching short videos, browsing news, listening to music and many other fields, are inseparable from algorithm recommendations.
However, in the past few years, algorithm recommendation has served the interests of enterprises, and there have been frequent problems such as riders trapped in algorithm systems, big data killing, and user privacy leakage.
The introduction of the new algorithm regulations refers to the problems of big data killing, algorithm discrimination, and inducing addiction, and together with the Anti-Unfair Competition Law, the Data Security Regulations, and the Personal Information Protection Law, it puts shackles on algorithm recommendations.
Some people call 2022 the first year of algorithm supervision, and algorithm recommendation will also usher in a new era.
History of algorithm recommendation
Algorithm recommendations were first applied overseas.
In 1998, Amazon launched an item-based collaborative filtering algorithm (ItemCF algorithm) to push the recommendation system to the scale of serving tens of millions of users and handling millions of goods.
The use of the ItemCF algorithm increased Amazon's sales by more than 100 times, quickly expanding from books to other categories and involving multiple industrial sectors, which was a successful application of algorithm recommendation from the laboratory to commercial companies.
Subsequently, more and more enterprises began to use algorithmic recommendations, and gradually emerged in the industry.
In 2006, the original DVD rental company Netflix Bounty Million Recruitment Algorithm Recommendation System, and with the results of this competition successfully transformed into an online video on-demand platform, at one time more than 80% of Netflix movie viewing came from the recommendation system; in 2010, YouTube announced that they used ItemCF to make video recommendations. Subsequently, social media such as Facebook and Twitter have also adopted personalized information streams, arranging content according to user interest.
Domestic Internet companies took the lead in embracing algorithm recommendation, and it is inseparable from one person, that is, Zhang Yiming, founder of ByteDance.
Zhang Yiming firmly believes that "personalized recommendation will inevitably become the trend of the future" and insists on the supremacy of algorithms. At that time, after he resigned as the CEO of Jiujiufang, he founded ByteDance to make today's headlines.
In 2012, the news client product recommended by the main content algorithm and replaced manual editing through machine learning was launched, and in less than 90 days, today's headlines had 10 million users.
At that time, the news field was still sina, Sohu, Netease, Tencent four major portals to fight to the death, almost no one put Zhang Yiming, a young man with no news experience, in his eyes. As everyone knows, Zhang Yiming is taking his algorithm to completely innovate the entire journalism industry and even the Internet industry.
Zhang Yiming once introduced the recommendation mechanism of today's headlines in his signed journal article "Machine Substitution Editing" - when the user binds weibo to log in within 5 seconds, the system will establish a DNA interest map for the user.
The weapon of success in today's headlines lies in the algorithm, and the secret of the success of various Apps under ByteDance also lies in the algorithm recommendation, which is the so-called "algorithmist has it, and the byte is particularly strong."
Another well-known product of ByteDance, Douyin, also relies on algorithms to accurately push interesting content for users, and in just a few years, it has become the national-level application that occupies the longest time used by Chinese netizens, and its unique algorithm also brings accurate conversion of traffic to advertisers.
According to the "2021 Douyin E-commerce Ecological Development Report" released by Douyin E-commerce and Giant Arithmetic, in January 2021, through the advantages of content + algorithm, the total amount of commodity transactions achieved by Douyin's interested e-commerce increased by 50 times year-on-year.
In 2022, today's headlines just celebrated its tenth anniversary, and today's ByteDance has successfully joined the ranks of Internet giants.
According to the "2021 Hurun U40 Young Entrepreneur List" released by the Hurun Research Institute, Zhang Yiming jumped to the top of the list with a wealth of 340 billion yuan, which is inseparable from Zhang Yiming's biggest outlet in the past decade - algorithm recommendation.
In addition, the rise of the new giant Meituan, Didi is also inseparable from the algorithm, according to the recruitment report, today's headlines are to take the recommended algorithm route, that is, to maximize the right content to match the right users, Meituan and Didi is based on LBS service matching, because this service is dynamic matching, so it needs a more complex algorithm system.
Nowadays, algorithm recommendations have spread all over the Internet, and Taobao, Zhihu, WeChat, Weibo, etc. all have sophisticated user preference algorithms to push users the content they are interested in.
Why should algorithm recommendations be restricted?
Internet giants have used algorithmic recommendations to develop rapidly over the past many years, achieving huge revenue and profits, and they have not only mastered the entrance to user traffic, but also mastered massive data based on user behavior.
Algorithmic recommendations are more like a double-edged sword as a product of the Internet age.
Compared with manual recommendation, the superiority of algorithm recommendation is self-evident, but the algorithm knows users better than the user himself, and the user under the algorithm recommendation is more like "running naked" on the Internet. Robin Li bluntly said in a public forum that Chinese consumers are willing to exchange certain personal data for convenient services.
In fact, society has appeared for many levels of "perceived discomfort" of algorithms, and the legitimate rights and interests of consumers have been seriously endangered.
On the Douban "Anti-Technology Dependence Group", many young people who want to fight against the algorithm are gathered, and they try to escape the algorithm and regain the initiative of life by downloading only communication apps on their mobile phones, watching videos and shopping to choose the web version, and carrying cash with them.
Anti-technology dependence panel discussion topic, photo/Douban
The new algorithm regulations jointly issued by the regulatory authorities refer to algorithm discrimination, "big data killing", inducing addiction, "choosing one of the two" and other phenomena, which are the adverse effects of the unreasonable application of algorithms, and these problems have long been drawbacks, which have been exposed many times before, but have not been completely solved.
Taking "big data killing" as an example, Internet companies use the user data they own to implement "price discrimination" and "price gouging" behavior against old users. According to financial magazines, big data killing first entered the public eye in 2017. At that time, some Weibo users posted that they were "killed by big data" by an online travel platform and an online ride-hailing platform, which quickly triggered public opinion and even became one of the top ten buzzwords of annual social life.
The Beijing Consumers Association disclosed a set of data in early 2019, and 56.92% of the respondents said that they had experienced being killed by big data, and as many as 88.3% of people believed that killing was a common phenomenon. Among them, online shopping, online travel, online car-hailing and other fields are high incidence areas of big data "killing".
At the end of December 2020, "big data killing" was once again on Weibo's hot search, this time the protagonist was Meituan. An article titled "I was cut leeks by Meituan members" on social media, the author of the article in the same store on the Meituan, with the same delivery address, at the same time to order, members than non-members of the delivery fee is higher, the article also said that I thought that opening a takeaway member will save money, but almost all the takeaway merchants nearby The delivery fee is 1 yuan to 5 yuan higher than that of non-members.
There are more and more problems with big data killing, and the regulatory authorities have already intervened. In April 2021, the Guangzhou Municipal Bureau of Market Supervision and the Municipal Bureau of Commerce intervened to hold a special investigation on the platform "big data killing" and an administrative guidance meeting to regulate the order of the fair competition market, and a total of 10 Internet companies, including Vipshop, JD.com, Meituan, Ele.me, Daily Excellent Fresh, Hema Fresh, Ctrip, Qunar.
The new algorithm regulations also clearly stipulate the problem of "big data killing", algorithm recommendation service providers to sell goods or provide services to consumers, should protect consumers' rights to fair trade, and must not use algorithms to implement unreasonable differential treatment and other illegal acts on transaction conditions such as transaction prices according to consumer preferences, trading habits and other characteristics.
Algorithms should not be just cold money-making tools for businesses. Sun Ping, deputy director of the World Media Research Center of the Institute of Journalism and Communication of the Chinese Academy of Social Sciences, said in an interview with Banyue that although there have been actions in the policy aspect, it still fails to effectively touch many problems of the universal scenario of algorithms. If you want to guide the rational development of algorithms, you should establish algorithmic ethics at the social level and create universal guidelines and ethical rules for the application of algorithms, so as to achieve the "common benefit" of human-computer interaction and the universal benefit of technology.
With the introduction of new algorithm regulations, which companies are affected?
The introduction of new algorithm regulations means that industries related to algorithm recommendation have begun to enter the era of strict supervision.
Previously, the regulator's penalties for algorithm recommendation mainly focused on monopoly and big data killing behavior.
On April 10, 2021, the State Administration for Market Regulation fined Ali 18.2 billion yuan for engaging in the monopoly of "choosing one of the two" in the field of e-commerce. After that, another Meituan was fined 3.4 billion yuan, the reason is also "two choice one", meituan in order to maintain its competitive advantage, many times forced merchants in the Meituan and other takeaway platforms to engage in "two choice one".
Not long ago, because of the problem of big data killing, Ctrip was sentenced to pay back one and lose three. According to Beijing Youth Daily, on December 31, 2021, the Shaoxing Intermediate People's Court rendered a final judgment in accordance with the law on the case of Hu X v. Shanghai Ctrip Commerce Co., Ltd. on infringement liability disputes, found that Ctrip Company constituted fraud, and ordered Ctrip to refund the difference in the reservation price of Hu X and pay compensation at three times the difference in room rate, which can be regarded as the success of the first case of "big data killing".
For platforms, the introduction of new algorithm regulations means that all kinds of Internet companies that provide algorithm recommendation services are almost within the scope of supervision, including various short video platforms, e-commerce platforms, social platforms and catering takeaway platforms.
It is worth noting that the algorithm provisions released this time add a new article to the draft for comments, requiring algorithm recommendation service providers to provide Internet news information services, they shall obtain service licenses in accordance with law, standardize the development of news information gathering and editing and publishing services, reprint services, and dissemination platform services, must not generate synthetic false news information, and must not disseminate news information released by units that are not within the scope of national regulations.
This also means that such platforms continue to provide Internet news information services, or need to obtain licenses and review the qualifications of information publishers.
After the implementation of the new algorithm regulations, it may compress the profit space and user growth rate of previously algorithm-driven enterprises to a certain extent.
When the growth rate of users slows down, it is foreseeable that the Internet industry will also fall into stock competition, and how to enhance user stickiness has become the key.
For users, the supervision of algorithm recommendation is essentially a good thing, so that the relevant industries grow from barbaric to have laws to follow, so that the rights and interests of users are protected accordingly, and the platform provides users with content and services that they really need.