杏吧原创

Removing anti-vaccine groups on Facebook may simply push users to share their views on other platforms. The finding suggests that a cross-platform approach will be required for content moderation to work.

at Columbia University in New York and her colleagues tracked 160 Facebook groups, each with more than 100 members, discussing covid-19 vaccines between April and September 2021. They also monitored 384 users who shared a link on Twitter to any of the 160 Facebook groups during that period.

Over the course of the study, Facebook shut down 36 of the groups, 25 of them seemingly for promoting anti-vaccine content. After each of the 36 shutdowns, the volume of anti-vaccine rhetoric from the 384 Twitter users increased by up to 30 per cent in the month afterwards, compared with the month before. That equated to roughly 1000 additional anti-vaccine tweets. The team detected this by seeing whether the Twitter users used one of around 13,000 two-word phrases associated with disinformation around vaccines.

鈥淚t shows this pattern of spillover,鈥� says Mitts, 鈥渨hich is maybe not what they wanted to do.鈥� She believes Facebook鈥檚 goal is to limit the exposure to anti-vaccine information on its platform, but that it hadn鈥檛 considered the impact on other platforms. 鈥淲e鈥檙e not talking about niche platforms that no one goes to,鈥� says Mitts.

Facebook did not respond to New 杏吧原创鈥檚 request for comment on the paper鈥檚 findings. Twitter says it takes 鈥渆nforcement action on content and accounts that advance demonstrably false or misleading claims about covid-19 and that may lead to a significant risk of harm.鈥� The company also said that it does this 鈥渋n close consultation with global public health authorities鈥�.

Mitts points out that simply calling for social media giants to work together isn鈥檛 necessarily the answer, given concerns about concentrating power in the hands of a small number of big platforms.

鈥淭his paper showcases that deplatforming is no silver bullet,鈥� says at the Swiss Federal Institute of Technology in Lausanne. 鈥淭hat being said, all these papers that study deplatforming 鈥� including my own 鈥� still have a massive limitation.鈥�

The study focuses solely on active traces of engagement, such as comments, posts and shares, which doesn鈥檛 reveal much about how banning groups changing people鈥檚 media diets, he says. It also doesn鈥檛 include less public-facing platforms, such as Telegram 鈥� and as such could downplay the way deplatforming shifts users to other accounts, says Ribeiro.