The Complex Relationship between Offline Events and Online Hate Speech

This new study analyzed data from 1150 online communities across several platforms, including Reddit, Gab, and 4chan, tracking the frequency of seven types of hate speech: racism, anti-Semitism, gender-based hate, hate against the LGBTQ+ community, general insults (E/I/N), hate against religion, and hate against immigration. The researchers found that racism was by far the most prevalent type of hate speech, appearing in nearly 7 million posts.

However, despite originating from hate communities, most posts did not contain any of the seven types of hate speech, and about 20% of hate posts contained more than one type of hate speech. The researchers also found that during periods of intense and contentious offline activity, the activity in online hate communities increases dramatically, likely due to the heightened salience of these events more generally.

The study found that different events trigger spikes in online hate speech, and these spikes vary by type of hate speech. The largest spikes occurred in January 2020 with respect to religion and anti-Semitism, February 2020 with respect to immigration, and May and November 2020 with respect to various types of hate speech. Structural break analysis confirmed that these changes in the data were not in line with prior patterns of hate speech and are the largest breaks in the time-series.

Interestingly, the study found that different events had varying effects on the relative frequency of hate speech within online communities. During the early days of the Black Lives Matter protests in May 2020, the percentage of posts that contained hate speech increased, meaning that hate posts increased not only in absolute terms but also in relative terms compared to other posts. However, during the period around the US elections in November 2020, the percentage of posts that contained hate speech decreased, even though there was an increase in hate posts in absolute terms.

The researchers also found that the mean number of hate speech types contained in hate posts remained quite consistent over time, with only two relatively small decreases occurring during the events mentioned above.