With regard to the second update, users covered by the greyscale test, after opening the subscription message page, find a page that would have shown a " one-and-a-half " subscription to a " one-and-a-half " subscription to a " one-and-a-half " subscription message, and other subscription messages are folded in " expand the earlier message " 。
These two updates reflect a line of thought on the part of the team: both the creators are expected to expand their output and the users are expected to expand their reading — both of which are in fact in the appearance of each other。
If the platform is planned to be enhancedAlgorithm recommendedDiluting the subscription mechanism will require ensuring adequate content availability; if the platform wants creators to be able to upgrade their capacity, it will need to provide a wide enough uplink for laterists. When the number of public ids per subscriber is almost saturated, this also means that uiw intends to break the “subscription monopoly” by making it available to more users through recommended algorithms。
While this logic is reasonable, it seems to run counter to the standards and concepts that the public platform once upheld. To put it straight, yesContent distributionInstitutionally, even with its own short-term headline. In any event, however, this is necessarily a change in the balance of pros and cons made by the team, the driving force behind which is the inevitable, but necessarily modified, oligoecology inspired by traditional distribution mechanisms。
Subscriptions lasted nine years. Where did the new live water come from
In 2019, zhang xiaolong gave a four-hour public lecture on twitter, reminiscent of the birth and design of several important functions. Speaking of the public, he said:
Texts, e-mails can be sent without control, and the side effects are known. So i think that if we provide a subscription-based model that avoids harassment and fraud of users and allows the service to send information to those who need it, it is a bridge between c and b. I remember being particularly excited when i thought about it, and telling ponny how powerful this mechanism would be. Ponny asked me, what about the trash? I mean, it doesn't have any junk。
As zhang xiaoyu originally envisaged, under the subscription system, users could never be harassed because 100% of the users chose to receive the message. This principle has been adhered to for nine years since the birth of the public sign in 2012, and in the same nine years, the headlines have grown and the companies behind them have become one of the biggest known opponents。
At some level, subscriptions and algorithms are two different paths to the same goal, both of which aim to bring the right content together with the right user, except for one that is more dependent on human choice and one that is more machine-dependent. These two product paths lead to social communities, while the other leads to the media。
Initially, the community of friends assumed the function of social distribution, and it was not until 2020 that wesson added algorithms for integrating social relationships to the "look and see", but still kept subscription streams for users that had broken the time order。
Zhang xiaolong originally planned that the source of traffic for the subscription number should be based on 2/8 principles: 20% of users select content in the subscription number and 80% of users read content in the circle of friends. That is to say, the circle of friends is the main entry point for users to find good content and good accounts, and the most central referral mechanism is that of acquaintances。
This logic is more in line with zhang xiaoyu's original vision of wissin. While familiar referral mechanisms are highly efficient when there are only dozens of like-minded good friends in the circle of friends, the attributes of the circle of friends will inevitably change with the almost unlimited expansion of the number of untrustworthy friends (it is hard to say that 5,000 of them are “know-how”)。
In 2020, nearly a million people had close to 5,000 friends in their tweets, and zhang xiaolong described his feelings with "quiet fear." at the same time, more than 100 million people set the circle of friends "just three days to see." one year later, that number rose to 200 million, representing a fifth of the twitter users who did not want their friends to see too much of their dynamics。
During the same period, the subscription interface was changed from a list to a stream of information, adding the word “good” (later replaced by a more emotional colour of “seeing”) and allowing the word “seeing” to act as an expanded distribution path。
It is clear, however, that it is not enough to distribute content based on subscriptions, which makes it difficult for users to gain access to content other than the list of concerns, and that the rise of backwarders is not broad enough, oligopolization becomes increasingly evident and there is no living water in the ecology. Can you recall, this year, what new public figures you've found and will read more often
Is this a big loss for users? This may not be significant in the short term. In terms of the current number of public figures and creators, users have full access to a sufficient number of content of good taste. The most affected were the late creators and the community itself, which in fact was the problem of “unequal distribution of dividends”。
Although zhang xiaolong had stated long ago that the public platform was not designed for the media but for “elimination of geographical constraints”, “elimination of intermediaries”, “connection of people and services”, the “summary (disappeared) was never within our consideration”. In fact, it is true that the most skilled and productive of content is the mass media industry, and the significance of its existence and its philosophy may have to change。
The problem of unequal distribution of dividends and ecological oligopolization has now become so serious that the mere inclusion of algorithms in the “look and see” recommendation cannot be resolved that it has had to shake the long-standing base of subscriptions. Does that hurt the wisps? Maybe we can wait for the wiss option after this test。
The possible impact of the algorithm-recommended mechanism on twitter
While the new public-intelligence mechanism is still being tested, we are already in a position to predict its possible impact once it is operational。
1. Platform advertising revenues have increased and creators ' advertising revenues are no longer stable
By referring to the headline system, algorithm recommendation + information flows can increase the supply of content and, of course, the stock of advertising and create more advertising spaces. The official advertising platform will have a very large voice, and the creators themselves will be weakened by the fact that the amount of reading is not guaranteed。
2. Content-to-user impact rate may decrease
Zhang xiaolong, explaining that the public number was changed from the list to the flow of information, stated that the number of contents in a screen should be inversely proportional to the hit rate。
The hit rate is the proportion of the number of content bars that the user may be interested in. For example, ten e-mails, only one is what i want to see, and the hit rate is 10%. Often, the more people focus on content, the lower the hit rate. So when the public sign was recast, the number of articles on a screen changed。
In this line of thinking, the more the system understands users, the less information a screen can be designed; rather, the less it is known about the user system, the better it is to give users more options。
Under the original subscription system, a screen of “one-and-a-half” messages, although small, was sent by accounts of interest to the user, so the hit rate increased and the user could also adjust the hit rate by removing the attention. But the effect of the current greyscale test is that a screen of "one subscription + half a recommended message" is actually risky, because about 50 percent of the probability is completely missed。
Moreover, it is important to note that the recommended model of a short video product, represented by tremors, “one screen short video”, is efficient but does not fit into textual content。
A well-known "gold six seconds" principle that most viewers can judge in six seconds that they don't like video content and cross it, so even if the impact rate on this screen is zero, the experience of users is not very damaging. But it often takes longer for the user to judge that he does not like a text, and if it takes two or three minutes to judge that he does not like, the damage to the user experience is serious。
3. The increase in content volume is accompanied by the fact that content quality is difficult to control
It is true that current algorithms are more susceptible to sensational titles, unnutrientful gossip and the trick of wiping edges than those recommended by acquaintances, because they represent the instinctal choices of most people. In an environment of semi-acquainted relationships, the introduction of a growing number of algorithms that are far from ethically mature would raise more social ethical issues and be highly doubtful。
An example of this by analogy is the “false-dong incident” in which a number of low-line urban and rural middle-aged older women users were misled by basic techniques such as tremor marketing using face-to-face combinations and exploited by those behind algorithms and algorithms, resulting in property and emotional losses, leading to many social problems. For all large dau products, users ' proficiency in the internet varies significantly, as does their exposure to algorithms. Even a product with a strong social community identity, such as tremors, can be plagued by this, with a stronger social identity that requires vigilance。
Moreover, to quote zhang xiaolong himself, “it is difficult for most people to write a text”. One argument is that one reason why the public domain preserves better quality than the content of platforms such as today's headlines, 100 houses and so forth is because of the limitation of “one day in a group”. Human creativity is limited, and when there is only one voice a day, there is less nonsense and more。
With the lifting of restrictions, opportunities have changed, but creativity cannot rise overnight. This must have been anticipated by the twitter team, so that the “publishing” function was separated from the “grouping” function during the test, so that users were not suddenly overwhelmed by bullshit. But don't forget that “published” articles can also be forwarded and appear in the recommended stream, so you always have the opportunity to see them in the dialogue box or in the circle of friends。
