While Twitter had previously suspended 40 percent of the primary 55 hate accounts, according to Bot Sentinel, the accounts were employing methods like including “parody” in their profile bios or using coded racist language to avoid penalty. In the process of conducting the investigation, Bot Sentinel used Twitter accounts without friends or followers, but that didn’t stop the platform’s algorithm from taking notice of their Twitter habits. “After viewing two hate accounts, Twitter’s algorithm began suggesting numerous hate accounts,” the report noted. “On multiple occasions, Twitter recommended we follow these hate accounts.”
As the report summarized, “Our research found that a relatively small number of single-purpose anti-Meghan and Harry accounts created and disseminated most of the hateful content on Twitter. However, the primary accounts had assistance that allowed their content to be repackaged and shared by accounts with a considerable following. We observed the primary accounts coordinating their activities and using various techniques to avoid detection. In short, the majority of the anti-Meghan and Harry activity wasn’t organic.”
Christopher Bouzy, founder of Bot Sentinel, attributed an immediate change they saw online to the report. “We observed a significant decrease in the activity targeting Meghan and Harry shortly after we published the report,” he tweeted. “If the activity had been organic, it is highly unlikely we would have witnessed such a sharp decrease in such a short period of time.”
He also wanted to make a distinction between “a group of accounts who just say ‘mean things'” and the accounts they were referencing. “They are single-purpose accounts dedicated to creating and disseminating hateful and often fictitious content meant to deceive,” Bouzy wrote, “and they openly coordinate on and off Twitter.”
