Internet: Twitter partly publishes recommendation algorithms

Internet: Twitter partly publishes recommendation algorithms

Twitter users should now be able to better understand the criteria according to which tweets are selected for them by the software. Elon Musk promises more transparency – and has to ask uncomfortable questions.

Twitter has largely made the algorithms used by software to select tweets for individual users public. The aim is to create “trust through transparency,” said Twitter owner Elon Musk in an online question and answer session over the weekend.

The removal of the free verification symbols, which was supposed to start on Saturday, was not initially noticeable. Various profiles of celebrities who did not want to pay for new subscriptions were still verified on Sunday.

Musk had announced the publication of the recommendation algorithms months ago – and Twitter finally made good on the promise on Friday. With Twitter, users can view tweets either in chronological order or selected by software. In the latter case, tweets from profiles that are not actually followed can also be selected in the “For you” area.

These recommendations take into account, among other things, which posts most recently aroused the interest of a user. The software also evaluates which tweets the profiles you follow interact with. Likes and retweets significantly help the distribution of posts, as experts found when analyzing the code. It’s not the full software that Twitter made public. It is unclear which parts are missing.

Musk said he thinks users shouldn’t trust any software algorithm on a platform that they can’t understand. When evaluating the software code of the algorithms, experts quickly discovered that there is a separate category for his tweets. When asked about this during the question and answer session, Musk called it “strange” and assured that he did not know about it.

Excitement about category “Elon”

A Twitter software developer said the “Elon” category only exists for statistical purposes and that there is “no preferential treatment in the algorithm”. According to the software code, another category in this area is whether a user is a Democrat or a Republican in the American party system. With the evaluation of such information, one wants to ensure that no groups are preferred or disadvantaged when changes are made to the software, said the Twitter developer. Musk announced that the categories would be removed quickly.

The emergence of the Musk category is the first evidence that Twitter is at least statistically tracking how well its owner’s tweets are performing. In mid-February, a report that Twitter temporarily increased the reach of Musk’s tweets also called supervisors in Germany into action. The Bavarian State Center for New Media (BLM) announced that it was investigating whether such manipulation had been observed on Twitter in Germany. Since then, there has been no information on the course of the investigation.

At the time, users noticed that among the contributions selected by the software in the “For you” area, there were suddenly a large number of tweet replies from Musk. The industry blog “Platformer” reported, citing Twitter employees, that developers of the online service had previously been asked to drastically increase the reach of Musk’s posts. The trigger was Musk’s annoyance that his tweet about the Super Bowl football final was displayed much less frequently than that of US President Joe Biden.

At the time, Musk denied that there was a deliberate increase in range. It was just a software error that made the answers weighted the same as tweets. At the same time, the algorithm actually contains precautions against too many tweets being displayed by one person.

Source: Stern

Leave a Reply

Your email address will not be published. Required fields are marked *

Latest Posts