Tik Tok doesn’t read minds, but the New York Times wants you to believe otherwise.
A December 5 article describes a leaked document obtained by the Times that reveals algorithms that are allegedly used to drive engagement on Tik Tok. While the Tik Tok algorithms have already been discussed, the article claims that the leaked document “offers a new level of detail on how the algorithm works.”
The article has several omissions. One of them is the lack of explanation of how and why the algorithm leads to a particular consumption of content. While the algorithm appears to rate videos based on parameters such as likes, comments, and watch time, there is no question which feature seeks to optimize the video, such as total engagement (hours spent), for example, or total delivery of the video. video. content, its “viral” quality.
Most telling is that the Times article uses misleading language common in media discussions of artificial intelligence, attributing things like “mind” and desire to what is just a simple technical feedback loop.
The title of the article, “How TikTok Reads Your Mind,” is followed by references to how the algorithm detects people’s intent:
The document brings a new level of detail to the video application, providing insightful insight into both the mathematical core of the application and the company’s understanding of human nature: our tendencies to boredom, our sensitivity. Cultural cues, which help explain why it is so difficult to understand. Let it go.
However, nothing about human nature is revealed in the discussion of the algorithm in question. The algorithm, based on the paper, appears to be a very simple factor calculation as follows:
Plike X Vlike + Pcomment X Vcomment + Eplaytime X Vplaytime + Pplay X Vplay
The expressions “like”, “comment”, “watch time” and “read or not” are probably references to the different parameters assigned to the videos. The article does not explain the “P”, the “V” or the “E”, although it is implied in the article that P represents a prediction “driven by machine learning”, without further precision.
As such, the algorithm summarizes the predicted metrics of the content, regardless of the human mindset.
Artificial intelligence company DeepLearning.AI, founded by researcher Andrew Ng, published the article in the company’s newsletter, TheBatch, last week. The article suggests that the “V” could stand for “value”, that is, a weight applied to each of the measures according to their importance in the final score.
Flowchart of a supposed recommendation algorithm used by Tik Tok, reproduced by The New York Times from an internal document.
Despite the omissions, it is clear that the system does not predict mindset, presumably it matches the pieces of content with the expected results in terms of possible views and / or participation.
Assuming that there is a mind within the user expressing their preferences by clicking, such as the article, is conjecture that may not be supported by the facts. In statistical terms, the fact that a machine can read a user’s mind would imply the notion of a “prior”: something that exists before a measurement. However, what is revealed as spirit, in the form of expressed preference, is actually the opposite statistical notion, a “posterior”, something that only exists after measurement.
It seems more likely that the mind is something inferred after the fact, if it makes sense at all. Consider the system that constitutes Tik Tok. Users can download and watch various short-form videos. As users submit videos and consume them, more such videos are presented to them. In a sea of videos, an individual clicks or does not click, engages or does not participate.
No matter what the mindset or emotions of the user are, they are irrelevant because the system does not ask for voluntary ideas. Rather, the user is asked to respond to a limited set of options, and the system is getting better and better, presumably to repeatedly stimulate this activity, leading to a growing number of daily active users which, according to the Times , is now on the order of a billion and is expected to reach 1.5 billion in 2022.
All of this suggests that Tik Tok is a highly efficient behavior machine at best, a behavior modeling machine on Tik Tok, rather than a mind-reading device.
Taking the analysis further, Tik Tok’s studies in the academic literature suggest a very mixed view of the algorithm at work.
In some cases, the company’s algorithm works not only to propagate things that might be popular, but also to expose things that might not be so popular.
For example, two Carnegie-Mellon researchers, Daniel Le Compte and Daniel Klug, this year interviewed activists who use Tik Tok to present videos aimed at drawing attention to social causes. They reported that activists expressed their preference for Tik Tok over other social networks because its videos were viewed more widely than on other platforms:
Some participants noted that using TikTok helps convey their message beyond their own ‘circle’: “Some participants noted that using TikTok helps them convey their message beyond their own. , uh, Facebook, where there are only friends of friends or family that you know in real life.
While Tik Tok can circulate things that go beyond what one person expresses a preference, it also seems true that Tik Tok activity focuses on things that groups of people approve of in large numbers, regardless of what an individual may. feel or think about them. .
A 2019 study by researchers from the Guilin University of Electronic Technology, China, and the University of Oslo, Norway, looked at the number of views and likes of Tik Tok videos. The authors concluded that most of what is published is what users like:
In particular, the number of views and the number of reviews have a very high correlation coefficient of 0.91, which means that a video that is popular in terms of number of views is likely to be popular in terms of number of views. ‘ and vice versa.
Again, continually showing users more and more things that they have already been shown and clicked on is a matter of technical feedback, not reading.
A third study, conducted this year by researchers at Boston University, Binghamton University and University College London, raises the question of whether Tik Tok’s recommendation “engine” is really doing anything. The study analyzed 400 Tik Tok videos in an attempt to “understand the metrics to make a short video go viral.”
The authors tagged the videos according to ten different factors that could affect virulence, or, as they call it, “virality,” that is, the propensity of a video to be liked by users. These factors ranged from the fact that the creator of the video was “popular,” meaning he had a large following, to the style of the video, such as using the “duets” feature in Tik Tok to remix someone. the dance floor of others and the emotional content, among other things.
The authors also sought to measure the role of the recommendation algorithm in conjunction with these other factors. They did this by noting how well the videos used relevant hashtags for their promotion and how long a video had been in the system, as viral videos tend to go viral shortly after being uploaded. Line.
The authors then used all of these factors in a variety of very simple statistical and machine learning models that can categorize things, including Random Forest, Support Vector Machines, Logistic Regression, Gaussian Bayesian, and Decision Trees.
The result ? Their classifiers, to varying degrees, were able to “identify the most important characteristics that distinguish viral videos from non-viral videos.” They found that the main factor is the popularity of the creator. The second most important factor was whether or not there were close-ups in the video, a finding that “matches previous studies of memes with images, suggesting that highly viral memes are more likely to use a close-up scale or medium plans. “.
So popularity increases popularity and people react to close-ups. None of this is mind reading. Meanwhile, they found that the recommendation system had the lowest value as a predictor of virality.
“RH2 characteristics (recommendation system) have the lowest AUC [aire sous la courbe] between the three HR [hypothèses de recherche], also possible that 0.71 “, écrivent-ils.” In fact, the precision obtained on these characteristics is legally possible (0.56), which suggests that they are not a good predictor of the virality of the a video.”
The authors also point anecdotally to the popularity of cat videos.
So studies suggest that Tik Tok may seek to impose certain videos on its users regardless of their desire or mindset, but that much of Tik Tok’s business is popularity contests and a herd mentality, quite obvious. None of this amounts to mind reading.
Rather, research suggests that Tik Tok can shape mental attitudes by reinforcing dominant tendencies in group behavior, for example by responding to popular “creators” who already dominate media consumption.
In other words, Tik Tok plays a bigger role in mindset building than mind reading.
Rather than speculate on mind reading, there are some fundamental aspects of social media that you should be aware of, including Tik Tok, that have nothing to do with spirits or mindset.
First of all, the machines are likely to take over social media activities. Watching videos and liking them are activities that fall well within software automation. Therefore, the idea that something must have a mind to participate is irrelevant.
Second, social media is a machine designed to produce a clear signal in noise. Individual preference, interest, or mindset has nothing to do with the purpose of the machine, which is to categorize behaviors into clear categories.
Finally, no one person has an identity or mindset on social media. What is called his personality, his mind, his identity, are only illusions, the consequence of a name attached to activities that are stored in a database.
For all intents and purposes, people don’t exist on social media, even though they spend time, a lot of time, using it. Therefore, no person, no spirit.
Woodmart Theme Nulled, WP Reset Pro, Newspaper 11.2, Newspaper – News & WooCommerce WordPress Theme, Premium Addons for Elementor, Rank Math Seo Pro Weadown, WeaPlay, WordPress Theme, Plugins, PHP Script, Jannah Nulled, Elementor Pro Weadown, Woocommerce Custom Product Ad, Business Consulting Nulled, Jnews 8.1.0 Nulled, Avada 7.4 Nulled, Nulledfire, Dokan Pro Nulled, Yoast Nulled, Flatsome Nulled, PW WooCommerce Gift Cards Pro Nulled, Astra Pro Nulled, Woodmart Theme Nulled, Slider Revolution Nulled, Wordfence Premium Nulled, Elementor Pro Weadown, Wpml Nulled, Consulting 6.1.4 Nulled, Fs Poster Plugin Nulled