Twitter periodically shares detailed information about the working logic of its algorithm. In the previously published analysis, it was revealed that the platform’s image cropping system applied negative discrimination to black citizens. The company admitted to the mistake, saying, “Bias in machine learning is a common problem and we are committed to improving it.”
Twitter has now shared its work on the status of political content within the application. In a blog post published on Thursday, he confirmed that his algorithm applies affirmative action to tweets posted by right-leaning individuals and institutions. He did not explain why.
Twitter examined 7 different countries within the scope of the study
Briefly touching on the functioning of the algorithm, Twitter gave information about its working logic. He reminded that since 2016, users can view tweets in reverse chronological order on the homepage, as well as list them in order of priority. He stated that in addition to the accounts you follow, he made suggestions that would be of interest to you based on the people you interact with frequently.
Twitter announced that it did the study to analyze the political content among the content displayed algorithmically on the home page. Underlining that it examined outstanding accounts in 7 countries, including Canada, France, Germany, Japan, Spain, the United Kingdom and the United States, between April 1 and August 15, 2020, the company detailed how the tweets from these accounts were reflected in the algorithm.
Stating that the political content shared by the accounts in which the review was made is highlighted in the algorithm, Twitter listed the findings as follows:
- Regardless of the political formation, the relevant tweets of our elected officials with political content were more prominent when compared to the political posts displayed in reverse chronological order.
- Our systems do not consider political ideology when making suggestions. For this reason, people who tended to the same political formation were not equally prominent.
- It was found that tweets shared by right-leaning accounts in other countries, except Germany, stand out more than left-leaning ones.
- It was found that right-leaning news organizations were more prominent by our algorithms compared to left-leaning news organizations.
Twitter admitted that it ‘reinforces certain political content’ as a result of the research. However, he did not share any information about the reason for this. Reminding that the algorithm takes into account the accounts that users interact with, he emphasized that it is difficult to determine the exact reason for the work in question.
41647
