““This column was written for several days, but no one read it” – I am sure you have repeatedly encountered similar headings, at least on socio -political themes on Facebook. It seems that a wave of bots, the rapid development of AI and the crisis of trust in social networks have already marked the point of non-return, and our favorite areas for online communication will no longer be as before.”, – WRITE: www.pravda.com.ua
“This column was written for several days, but no one read it”-I am sure you have repeatedly encountered similar headings, at least on socio-political themes on Facebook. It seems that a wave of bots, the rapid development of AI and the crisis of trust in social networks have already marked the point of non-return, and our favorite areas for online communication will no longer be as before. Let’s try to recognize hidden threats to Ukrainian users through simple content analysis.
The non -return point can be caused by current news. Recently, Mark Zuckerberg announced changes in the content of Facebook content. One possible solution is the use of moderation by users (the so -called Community Notes) by analogy with X, when users have independently added notes that the post could contain false information.
At first glance, it can solve the problem of blocking content on military topics, which currently exists and prevents Ukrainians from effective to hold a meeting, to tell about Russia’s military crimes and to inform the world about the situation in Ukraine. However, there are risks that are likely to harm and increase the spread of enemy misinformation.
The content of content moderation on Facebook is a very important thing. In our work, we regularly launch national campaigns of national level on all social networks. Facebook always stands out among others a large wave of irrelevant comments that are completely different from those that occur in other networks on a similar unit of content. The experience of my colleagues from other countries confirms the thesis that it is often the advertising campaigns on Facebook that attract the attention of bots.
First of all, let’s analyze the sentiment of comments under the publications of the most popular Ukrainian media. Let’s try to find similar topics, semantics, profiles and distribution of narratives that can be counted as artificial modeling and impact on the audience. Take the pages of “Ukrainian Truth”, liga.net, NV and “24 channel” on Facebook and compare the comments to their posts in the same period of time and similar topics.
In total, 110 comments were analyzed. There is criticism, repetitive topics and emotional colors, but comments are quite diverse. There are no clear groups that have a high text match (> 80% by TF-IDF analysis). This means that users are likely to write unique comments, not copy the prepared messages that would be a typical sign of botoferms. More than 90% of comments have a negative color and often do not match the news.
Instead, in content, they express distrust of power and violate the topic of corruption, express disappointment and pessimism, criticize the decisions of military leadership. Words that happen often describe actions related to political activity (“took”, “betray”, “sabotage”, “checked”), and express valuation judgments (“no”, “rejoice”, “rakes”).
Therefore, quantitative indicators can be assumed that there are comments from real persons, because there are no signs and patterns. Society is tired, disappointed and does not trust the state and, as a consequence, constantly criticizes it. However, if you resort to a more detailed analysis of users’ profiles who left these comments, new facts appear:
A large number of atypical, and actually non -existent names, lack of quality pages, but with excessive activity in the comments under the publications of Ukrainian media and reports with sharply negative criticism of the actions of Ukraine’s representatives, even if it does not relate to the topic Artificial modeling of narratives rooted firmly in the communication space. They can be detected, first of all, by poor quality profiles. As a rule, fictional names, lack of real avatar, participation in Arab communities, etc. are drawn. This is something that is easy to see at first glance. But quality bots are very difficult to distinguish from real people.
Community Notes is a potentially useful tool, but it risks becoming double weapons if “misinformation reporters” themselves turn out to be bots and social networks do not anticipate clear mechanics of combating them. After all, the bottle farms and artificial modeling of narratives have not disappeared. Some of them have obvious features, such as strange names, empty profiles and “collecting salt” with foreign subscribers. In the most popular information flows, bots are sinking in general discussion, but in small communities they can create a curved information picture. The question “who will believe” is increasingly relevant. The answer to it depends not only on Facebook algorithms, but also on each of us.
Yaroslav Makarevich, Action Global Communications in Ukraine