Source 丨 Pin Play(ID:pinwancool)
Author 丨 Guo Haiwei
"People just upload the data, and I don't know why. They 'trust me', a bunch of dumb fucks. ”
Source: The New Yorker
In 2004, a college student named Zach was starting a social networking site called Facebook. And the "sheer volume" of information is making the 24-year-old sophomore a little confused. Among the 4,000 messages, there may be the address of the girl in the next class, the photo of the campus star and the SNS. This classic dialogue, which metaphorically depicts the fate of social empires, was later written into a lengthy Facebook story by a reporter from The New Yorker.
Eighteen years later, the 4,000 private messages in Zack's computer quickly became nearly 4 billion user data stored in countless servers around the world. The latter underpins the social empire, earning more than $86 billion (548.1 billion yuan) of advertising revenue each year.
Apparently, Zuckerberg can no longer look through the information of 4 billion accounts like he did with 4,000 pieces of information. Instead, the tools that help him "penetrate" the information are called "algorithms." Not only social networking sites, but also around the world, whether it is search, shopping, news, short videos, where there is massive data, there are algorithms.
For a long time, "algorithms" have been the best assistants for giants. He helps Internet giants understand users, extend the time users spend in their content communities, and form massive and accurate business delivery decisions.
According to statistics, only Apple IDFA adjustment, the revenue of the advertising alliance will correspondingly drop by 30%-50%, or even more; and for the algorithm news platform, if there is no personalized content presentation after the algorithm support, the use time and active users of the existing algorithm platform will face great challenges.
This may have been unexpected by the sages of the Internet: we thought we were just giving up our attention and time for the platform, but we did not expect that we were also betrayed by our own information profiles. And even if we know the fact that our information is sold, we don't quite understand how this information is used by the platform.
In a sense, it is the 4,000 "silly Xs" who upload all kinds of private information for Zuckerberg free of charge to help Zuckerberg open up every night of money sleeplessness for the next ten years.
We are indeed the stupid X in Zuckerberg's mouth. Sadly, we didn't have a choice either.
Image source: PEXELS
On the one hand, there is a huge chain of interests, and on the other hand, there are users who lack the right to know and veto. Algorithms and the data specifications behind them have become the focus of global data regulators. On the first working day of the new year, China's algorithm management norms stood on a new level.
On January 3, 2022, China's "Provisions on the Recommendation and Administration of Internet Information Service Algorithms" was officially released, and the latter will be officially implemented from March 1.
This also marks that a few months after the release of the draft for comments, China's first rules and regulations specifically for algorithm recommendations have officially landed.
From the perspective of actual content, the draft for comments is slightly different from the official manuscript. In terms of regulations, the official version of the regulations adds more details, including the rights and interests of the elderly, and raises the fine standards of the corresponding regulations.
However, from the overall point of view, the "Provisions" basically follow the structure and content of the previous opinion draft, and all the key contents of network concern such as "anti-big data killing", "user modification of algorithm labels", "anti-unfair competition", "juvenile protection", "anti-addiction" and so on are all used.
In the official version of the provisions, the legislator also divided the original provisions into six sections, including "General Provisions", "Information Service Specifications", "Protection of User Rights and Interests", "Supervision and Management", "Legal Liability", and "Supplementary Provisions". The content of these different subjects also means a major change in the power relationship of the "enterprise-government-user" triangle of algorithms:
Source: The Cyberspace Administration of China | China's first regulations specifically for algorithms
"Algorithm neutrality" and "algorithmic innocence" will become history
In the past, under the principle of presumption of innocence, algorithm providers often had problems such as "unclear rights and responsibilities" in specific scenarios. The algorithm has easily become a synonym for nihilistic power - whether it is the traffic distribution of the B-side, the takeaway order; or the content reading experience of the C-end, the dynamic price system, the algorithm has the "highest interpretation power" of all the "unclear and unclear" negative results.
But the mechanism of the algorithm is often a "metaphysics" for ordinary users, using black boxes to explain black boxes, and many things can only be solved. Of course, compared with the vague power and responsibility space brought by the algorithm, the commercial benefits of the "certainty" behind the algorithm have returned to the hands of the platform.
The Provisions clearly state that the algorithm provider needs to open the black box behind the algorithm to the user:
"Users should be informed of their provision of algorithm recommendation services in a conspicuous manner, and the basic principles, purposes and main operating mechanisms of algorithm recommendation services should be publicized in an appropriate manner." "(Encourage) optimization of the transparency and interpretability of rules for retrieval, sorting, selection, pushing, presentation, etc.")
In terms of the purpose and use of algorithms, the Provisions also have clear guidance. For the purpose of the algorithm, in addition to not violating the existing national laws and regulations, the algorithm can not be used for "unfair competition", "inducing users to indulge", "excessive consumption" and other purposes, and it must "adhere to the mainstream value orientation" and "actively spread positive energy".
In the process of nodes, the Provisions propose that "the ecological management of the algorithm recommendation service page page should be strengthened, and a mechanism for manual intervention and user self-selection should be established and improved", and "illegal and bad information keywords shall not be recorded in the user's points of interest or used as user tags and pushed information accordingly".
This is tantamount to a warning of the potential derivative consequences of the original "algorithmic innocence" and "technology neutrality". Algorithms have tendencies and values in actual Internet practice, and the owners of algorithms need to bear responsibility for the values and content derived from algorithms.
Users will have more fine-grained algorithmic powers
On the one hand, the Provisions give users the power to "reject algorithms".
The Provisions require algorithm providers to "provide users with options that are not tailored to their personal characteristics" or "provide a convenient option to turn off algorithm recommendation services". Once the user performs the shutdown, "the algorithm recommendation service provider should immediately stop providing related services." ”
On the other hand, the Provisions also give users the power to "improve algorithms".
The Provisions require that "algorithm recommendation service providers shall provide users with the function of selecting or deleting user tags for their personal characteristics used for algorithmic recommendation services." "As users, we can also better understand the "tags" given to us by the system and make targeted deletion choices.
Personalized advertising label management on mobile Taobao
In addition, the Provisions also require algorithm recommendation service providers to give users more convenient form of supervision powers, "set up convenient and effective user complaints and public complaints, report portals, clarify the processing process and feedback time limit, and timely accept, process and feedback the results of processing." ”
In addition to the above codes of conduct in the direction of algorithms, the Provisions also "name" and encourage the positive role of some "algorithm strategies" in the construction of platform ecology.
For example, "encourage algorithm recommendation service providers to comprehensively use strategies such as content deduplication and fragmentation intervention, and optimize the transparency and interpretability of rules such as retrieval, sorting, selection, push, and display, so as to avoid adverse impacts on users and prevent and reduce disputes and disputes." ”
Among them, it is also content deduplication, fragmentation intervention, etc., which is consistent with the platform's long-term strategies of supporting the waist ecology and encouraging original authors, which is conducive to the development of the overall algorithm platform ecology.
The question behind all the norms is, who should be the master of the algorithm?
The algorithm appears to be programmer's code. But the reason why good algorithms can be established is more from user data and behavioral feedback - without data, there is no algorithm. In a way, the cornerstone behind the algorithm is the machine's "understanding" of each user and its behavior. On the basis of understanding, we make a probabilistic prediction of future behavior.
In the past, this right of interpretation and the right of earnings belonged to large corporations. How big companies "understand" me and my decisions used to be a "black box" for myself. So much so that most users never think about what it would be like if one day they could harness the algorithm.
The easiest way to exercise power is that I can turn away algorithm providers from profiting by misusing their own information.
And the deeper impact may be that in the future, I will soon be able to see my portraits on various platforms - the original B station thinks I like ghost animals, Douyin thinks I love to watch social hotspots...
And if the future platform can really make the rule mechanism transparent, can cede the power of the algorithm to users, and then allow users to decide their own content preferences. So, can I deserve an algorithmic content product that I prefer, which is out of simple and low-level fun?
This may not be possible as quickly, but at least one thing is certain:
Taking 2022 as the starting point, the "artifact" of the algorithm should belong to the people.
*Header image source: pexels
This article is reproduced from Pinwancool (ID: pinwancool) with permission, if you need to reprint it for a second time, please contact the original author