Authority and Network Effects: Twitter’s Algorithm
https://www.economist.com/graphic-detail/2021/11/13/according-to-twitter-twitters-algorithm-favours-conservatives
Twitter recently released information about what kind of bias their algorithm holds. In 2016, Twitter changed its feed operations. Previously, users’ feeds would show tweets of accounts they followed in reverse chronological order. After 2016, Twitter developed an algorithm that boosts relevant and designated tweets per user. Twitter kept 1% of users on the old reverse-chronological ordering of feed. This allowed them to perform an analysis of their algorithm and understand if it held bias. Initially, they found conservative-leaning tweets to be boosted at a greater rate than left-leaning tweets. After greater analysis from The Economist, it was found that the algorithmic bias was the result of a boost bias on the accuracy, as opposed to the ideology. The article identified that tweets that linked to websites with lesser accuracy, such as The Nation and the Daily Wire, were boosted at a greater rate than tweets linking to reliable sources like the Wall Street Journal and New York Times. This was irrelevant to ideologies, as unreliable sources from both sides were boosted. The level of reliability of the article was detected through an independent group: Ad Fontes Media.
Ad Fontes Media reveals bias in the media and gives rankings of reliability of sources. Ad Fontes Media is a hub, which gives authority scores to web pages. Ad Fontes Media has a high hub score, as the link to webpages with high authority scores. This does create a sort of cascade, as Ad Fontes Media links to more authoritative sources, they become a better and better hub. This is also true of the authorities that are linked. For example, Wall Street Journal becomes more authoritative the more Ad Fontes Media links to it, as Ad Fontes Media is a hub with a high score. The concept of tweets being boosted is also an interesting example of the Rich get Richer phenomenon. If a tweet has been boosted by the algorithm, this is because the tweet has received a great deal of interaction (retweets, comments, likes). However, the tweet has also been boosted because it has received lots of interaction. This is a circular argument because those that are already “rich” in interactions, will only become more “rich” as the algorithm continues to boost them.