How Algorithms Shape Our Lives and Relationships by Making Some Content Visible and Others Invisible

  Algorithms can almost be said to be the secret laws of our lives.
  As long as you open your mobile phone and read content, Toutiao, Douyin, Kuaishou, Xiaohongshu, Weibo, Bilibili, WeChat… it can be said that algorithms are everywhere. In addition to content distribution platforms, even daily life software such as takeout, taxi hailing, and grocery shopping are also deeply related to algorithms.
  Although we are all familiar with this word, what exactly is “algorithm”? Today, we will deeply explore the meaning and impact of this term from the perspective of media and communication studies.
What is an algorithm?

  First, what exactly is an algorithm? In fact, long before the invention of computers, the concept of “algorithm” existed. As the name suggests, it is a calculation using written calculation methods such as addition, subtraction, multiplication and division. After the advent of computers, algorithms specifically refer to a set of strategic mechanisms for solving problems in a systematic way. Simply put, it is a set of programs that automatically run using commands that the computer can execute.
  For example, using a search engine, after entering a keyword, there will be a set of calculation methods to rank web pages that have a certain relationship with the keyword according to relevance. This involves different ranking calculation strategies. Should we rank the web pages with the most visits first, the web pages with the most citations, or the web pages with the most repeated keywords?
  So the algorithm itself has a certain degree of subjectivity, and it shows people’s different understanding and value prioritization of solving problems. For example, a short video website will give priority to recommending the most popular videos to users, or giving priority to recommending new videos to users, whether to give priority to recommending completely similar videos, or to giving priority to recommending videos that are similar but with certain differences. This depends on the platform. Values ​​and Judgment.
  There are certain differences in algorithm recommendation mechanisms between Douyin and Kuaishou. For example, they will put newly released videos into a pool and give them the same opportunity to test. If a video has a lot of clicks, Douyin will always recommend it and keep it in the pool, making the video even more popular. This will form the Matthew effect, and popular videos will become even more popular.
  But Kuaishou is more “inclusive”. After a certain period of time, the videos in this pool will be replaced. In this way, although a video is very popular, it may not be continuously recommended, but will take appropriate care of some less popular videos.
  In other words, on Douyin, you are more likely to see videos that everyone is watching, while on Kuaishou, you are more likely to see videos that are different from what others see.
  According to some previous calculations, Kuaishou’s head content traffic accounts for about 30%, while Douyin’s head content traffic accounts for 88%. This data is not accurate and is just a rough estimate, but we can basically see the difference in traffic distribution policies between the two.
The Essence of Algorithms: Social Dimensions of Visibility

  So what is the essence of the algorithm? The essence of this is not a technical level, but a social level. Because we are not learning algorithmic technology, but understanding its cultural significance. From a social perspective, the essence of algorithms is visibility.
  One of the characteristics of media is to make the invisible visible. Algorithms, as a medium, also have this characteristic. They make certain contents visible and, conversely, make certain contents invisible.
  For example, on Weibo, the content updates of the accounts you follow are not arranged in chronological order. In other words, it is not that the most recently updated content is ranked at the top. The accounts you often watch, or interact with (likes, forwards, comments, private messages) or accounts that are relatively close to you on the Weibo social network will be ranked first every time. Open it and it will appear at the top of the page where it is most easily seen.
  On the other hand, those accounts that you don’t read and interact with very much will be ranked further and further behind you over time, and gradually disappear from your field of vision. Algorithms make some users or content visible and others invisible.
  From this point of view, do algorithms also play a role in constructing the world and our interpersonal relationships? Moreover, as a media or communication infrastructure, algorithms themselves are also transparent. When ordinary users use it, it is difficult to realize that the values ​​​​and framework behind the algorithm are affecting their views of the world and their relationships with others.
  So how does the algorithm work? It will first secretly collect data on user behavior, and then put this data into its calculation formula. This calculation formula is very complicated. What we often see published is just a very simple formula, but in fact the formula used may be more complicated. It is a trade secret and it is also a black box.
  This black box is not only mysterious to users, but also difficult to truly understand for the platform. Because it involves too many parameters and data, engineers can only set initial conditions, but it is difficult to control the actual precise results. Therefore, often affecting the whole body, adjusting one place may cause unexpected results in other aspects.
  Some scholars regard algorithmic rule as a Foucault-style panopticon, which was originally an ideal prison building envisioned by Bentham. The prisoners’ cells were arranged in a circle, with a surveillance tower in the center. People in the tower can clearly see what the prisoners in each room are doing, but the prisoners cannot see what is going on in the tower, and the prisoners cannot communicate with each other. In this way, prisoners will feel that they may be monitored at any time and obey the rules. Therefore, it is possible to monitor as many prisoners as possible with the least manpower.
  Social media algorithms have also constructed a similar panopticon. Users know that the algorithm exists, but they don’t know exactly how it works. If they abide by the rules of the platform, their information will be recommended and have visibility. Conversely, if they do not abide by the rules of the platform, they will be punished and will not have visibility.
Public domain traffic and private domain traffic

  The visibility mentioned here is mainly reflected in traffic. Generally speaking, traffic is divided into public domain traffic and private domain traffic.
  The so-called public domain traffic refers to the traffic obtained from recommended or selected content pages on the platform’s homepage. The recommendations you see on the homepage of Station B are public domain traffic. The opportunities for traffic distribution are very scarce. Generally, content that the platform considers very valuable will be promoted in this way.
  On short video websites, the videos or live broadcasts that will be recommended in prominent locations as soon as you open them are often public domain traffic. Because it is compulsory to a certain extent, this traffic is very considerable.
  Private domain traffic is the traffic generated by each account subscription or fan attention. That is, the platform does not recommend it, and some loyal users will actively watch the generated traffic.
  This type of traffic platform generally has relatively small control rights and is the basic account of the account. Of course, if the rules of the platform are violated, it can also intervene by limiting traffic or even banning accounts. Even if content review is sometimes delayed, the best time for publication may have passed.
  There are also differences between Douyin and Kuaishou in the distribution of public domain traffic and private domain traffic. For example, if an account posts a new video, Douyin will recommend it more to users who are interested in this content rather than to fans who have subscribed to this account. Kuaishou, on the other hand, will recommend more to fans.
  The traffic ratio allocated to Douyin users and fans with interested tags is 9:1, while Kuaishou’s is 3:2. The result of this is that on Douyin, no matter how many fans you have before, if the quality of a single video you post is not good, the traffic may not be too high. But on Kuaishou, as long as you have more fans, each video will have at least a relatively high guaranteed traffic.
  From this point of view, Douyin seems more fair and gives good videos a chance, but for big Vs, they cannot rest on their previous achievements and survive peacefully. If they cannot innovate, big Vs will soon be eliminated. disuse.
How does the platform’s allocation of traffic constitute a discipline?

  In addition to the subjectivity and discipline of the algorithm itself, the platform actually also holds a large amount of flexible traffic for distribution. If the platform organizes some activities and creators actively participate, they will be allocated a certain amount of traffic.
  Short video websites also have a certain amount of traffic during live broadcasts. The platform will assign some promoted anchors to some users who will enter the app by default when they open the app. In this way, the live broadcast will become very popular, and this popularity will bring more people to watch and follow the crowd. At this time, anchors with background and obedience will be rewarded, which also constitutes a kind of platform discipline.
  Even today’s platforms will commercialize traffic. Douyin has a DOU+ platform, and users can spend money to add traffic to their content. You can also buy views on Sina Weibo to make your content more visible. In addition to the platform, an underground market for traffic commoditization has also emerged. This is the “Internet troll army” that everyone has heard of.
  You can pay to get more viewers, fans, reposts and comments, and you can even use fake orders to create false sales or positive reviews for your products.
  The final goal of this kind of data work is still to meet the requirements of the platform algorithm and improve account visibility. Of course, if you buy a lot of traffic during live broadcast, but the things are not sold, can you make back the traffic fee you invested? This has certain risks.
Is it possible to “reverse engineer” an algorithm?

  As mentioned before, the algorithm is a black box, which leads to many “reverse cracking” operations. But this kind of “reverse cracking” is full of metaphysics. Because algorithms are composed of complex parameters and data, the human brain has no way to truly understand them. It can only make inferences and establish causal relationships based on input and output results.
  This is like Plato’s allegory of the cave. The tied slaves saw the projections of different objects passing by on the wall every day, and began to guess the patterns. However, there is actually no pattern in the appearance of these different objects. You may hit the right spot and think you have discovered the truth.
  Plato originally wanted to ridicule what he considered “civil science” here, but through this method, slaves also gained explanation and meaning of the world.
  Users’ speculation and explanation of algorithms is also such a process. Some scholars call it “algorithmic imagination” or “algorithm decoding”, which is how people use human meaning to understand the logic of the algorithm. In this process, there are still differences between users. There is communication and collaboration, a process some call “algorithmic gossip.”
  Looking at these concepts, you will find that scholars believe that these behaviors are not reliable. These algorithmic imaginations may make sense, or they may succeed by chance, but there is no guarantee that they are scientific.
  Because these so-called “algorithm experts” do not have a scientific method and only use mythical explanations, such as viewing algorithms as human-like existences. Some creators will summarize some methods to improve recommendation by the platform and increase traffic, such as continuously posting the same vertical content for a period of time, or increasing the duration of live broadcasts.
  In fact, these methods often use human thinking to speculate on algorithms. But this does not prevent people from exploring and giving certain meaning to algorithms. There are even some “algorithm experts” holding classes to explain their own experiences.
  In addition to imagining and complying with the algorithm, users will also resist, subvert or even rewrite the algorithm’s intentions, which is also called “playing the algorithm.” For example, some users will intentionally watch content they don’t like to interfere with the algorithm’s recommendation mechanism and break the information cocoon.
  In short, technology and algorithms themselves are now involved in the content production process, just like the professional ideologies, practices, macro-systems and other factors discussed previously, and user behavior and content creation are also disciplined by algorithms.
  Platform capitalism uses a seemingly neutral technology to more closely control user behavior, and users are actively adapting to and resisting this system. In a sense, the algorithm, as an actor, also participates in the production of content, exacerbating the commodification of content.