Detecting hate levels across social media

Juan Carlos Pereira Kohatsu, a 24-year-old data scientist, developed an algorithm that detects hate levels across Twitter. The Spanish National Bureau for the Fight against Hate Crimes, an office of the Ministry of the Interior, co-developed the tool and hopes to deploy it to ‘react against local outbursts of hatred’ (Algorithm Watch 2019). According to “El País” (Colomé, 2018), the algorithm tracks about 6 million tweets in 24 hours, filtering more than 500 words linked to insults, sensitive topics, and groups that frequently suffer hate crimes. Pereira’s analysis suggests that the number of hateful tweets remains relatively stable day-to-day in Spain, ranging between 3,000 and 4,000 tweets a day.

Public authorities are still figuring out the practical purposes behind implementing the tool. And, questions regarding the algorithm’s method of training remain unresolved. Until now, the algorithm has learned to classify tweets between hateful and non-hateful tweets according to subjective reporting. This could have the consequence of making the algorithm perceive hatefulness according to a specific sociodemographic group, which may not be representative of what is legally considered to be hate speech or at least what is not socially regarded as such by the majority of people.