Bad news for hateful commentators

Tomas Krilavičius, Dean of the Faculty of Informatics at Vytautas Magnus University, and Goda Gavėnaitė, Technology evangelist at VMU Faculty of Informatics

In April 2021, a virtual patrol was launched on Lithuanian Facebook - a great and undoubtedly necessary initiative to control the violations taking place on the platform. From the beginning of the patrol to November, over 2,700 reports of misconduct were received. Such a number would probably not surprise anyone who spends at least a few hours a day scrolling on Facebook. It is an open secret that we can easily notice at least one such violator ourselves. Every day, a variety of unclear content appears on this platform, from hateful comments to get-rich-quick schemes, which people are slowly beginning to recognize as scam. When such a case is detected, a virtual patrol can be informed, who in turn will start with a warning, but other inconveniences, such as pre-trial investigations, may occur later on. In the seven months of Facebook monitoring, as many as 77 cases were referred for clarification. The TOP 3 violations look like this: 28 cases of fraud are in the first place, illegal trade is the second with 18 "points", and incitement to hatred, with 15 cases in total, takes the honorable third place. Is it a lot or a little? How many remain undetected? What does this say about us (or them)?

Facebook's comment feature has been around since the platform's inception and is unlikely to be removed. We might say that maybe it's a good thing - people have the opportunity to spill their grievances in a virtual environment instead of elsewhere. Unfortunately, it does not come down to only one social network - rather, it is typical in the online space in general: harsh, inexplicably angry, and unacceptable language can be found on almost every portal, especially news portals. Of course, there is freedom of speech - it is impossible to ban commenting, but an even more unrealistic scenario is to wait for change in such a vast global network as Facebook. Freedom of speech on the Internet has brought people to a dangerous point, where a balance is struck every day between cute emojis next to photos and threats to get rid of "that gay", wishes to die and threats to kill. Such "self-expression" is a major societal problem that can be noticed with a few clicks on the screen. So, how do we solve it?

We should start with "ourselves" - for example, our Lithuanian news portals. VMU Faculty of Informatics research team - prof. Tomas Krilavičius, dr. Danguolė Kalinauskaitė, dr. Justina Mandravickaitė, Eglė Kankevičiūtė and Milita Songailaitė - will have something to offer in 2022 regarding this question as the project Be Hate-Free is developing an innovative tool to reduce the manifestations of hate speech on the Internet. At the same time, the initiative aims to address other related issues: raising awareness of hate speech and developing the public's ability to recognize it and the habit to report such incidents. To this end, the team of scientists join forces with artificial intelligence (AI) - the aforementioned tool, based precisely on AI, will help identify inappropriate, offensive and hateful comments. This innovation is being developed on the basis of comments posted by anonymous users on news portals, and the project alone includes almost 75,000 comments. Some of them are, of course, harmless, such as different opinions, but others are difficult to even comprehend, sometimes baffling and horrifying. All of those comments are carefully selected - annotated - manually. Generally speaking, they are categorized according to whether or not they contain hate speech. Artificial intelligence is learning from these examples so that later, when the tool is developed, it could carry out this work that is important to society on its own. This means bad news for hateful commentators as they will have to think harder before lashing out if they want their comments to bypass AI detection. 

The Be Hate-Free project team at VMU also includes students helping researchers and experts turn the initiative into a functional tool.  Young researchers have the opportunity to apply their theoretical knowledge to real practical problems and at the same time contribute to the noble mission of making the comments section a more peaceful and cleaner place. "I joined the Be Hate-Free project and at the same time I had the opportunity to write a master's thesis on the topic ‘Hate Speech Detection Using Artificial Intelligence.’ In this project, I contribute to research aimed at finding out the best neural network architectures and transformer models suitable for detecting hate speech in Lithuanian texts. My further goal is to work with the team to create a prototype of a solution for hate speech detection in Lithuanian texts, using machine learning methods," said Eglė, a student at VMU Faculty of Informatics working on the project.  Milita, a student at VMU Faculty of Informatics, who is responsible for the development of models for the detection of hate speech, as well as the collection and preparation of data for the Be Hate-Free project, said that at the beginning of the project, the existing methods for hate speech detection were tested, and the results were presented at the conference “Data Analysis Methods for Software Systems” (DAMSS) held in Druskininkai. Currently, the student is working on the development of a new model of artificial intelligence for the detection of hate speech in Lithuanian texts.  The AI-based tool is expected to be completed and launched this year.