Why banning hate groups online can accidentally make things worse

Using a model borrowed from statistical physics — where online communities coalesce and fragment like clusters in a fluid — researchers mapped how hate content spreads across platforms. They found the system has 're-entrant' phases: as regulators shut down hate communities, spreading first stops, but if crackdowns go too far without targeting the right nodes, it can restart. The result is a concrete warning that blanket 'do more' moderation policies can backfire, and the math points to specific thresholds where intervention actually works.