Study: simple peer contagion drove scholars' X-to-Bluesky migration

A new arxiv preprint, "Simple contagion drives population-scale platform migration", takes the question of how people actually leave a social network and answers it with the kind of data you rarely get for this sort of thing. The authors linked 276,431 scholars on Twitter/X to their profiles among the universe of 16.7 million Bluesky accounts, tracking them from January 2023 through December 2024, and used Brazil's court-ordered suspension of Twitter/X as an exogenous shock to identify causal effects.

The headline result cuts against a lot of the existing literature. The standard story for behavior that spreads through a network, especially something as high-friction as quitting a platform you have invested years in, is 'complex contagion', meaning you need reinforcement from multiple peers before you act. This paper argues the opposite for platform migration: adoption is peer-driven, but the mechanism is simple, not complex. One exposure does real work. The authors also report that treatment effects from the Brazil shock are short-lived and dose-dependent, and that the people most likely to migrate are those already deeply embedded in Twitter's social graph rather than peripheral users.

The retention finding is the part I would underline if you build or study these products. Early reconnection with prior contacts predicts longer tenure and engagement on Bluesky. That reframes the migration question from 'how do we get them to sign up' to 'how fast can they rebuild the graph that made the old platform sticky in the first place'. The authors also note that public political expression predicts migration, consistent with homophilous inflows into what they describe as a largely left-of-center Bluesky information space.

The honest caveat is that this is one population (scholars), one shock (Brazil), and one preprint that has not yet been through peer review. The cross-platform matching pipeline is described as high-precision but the paper is where you would go to interrogate that, and the abstract does not give error rates. What the reporting doesn't give you is how durable any of this is past December 2024, or whether the same simple-contagion finding holds for non-scholarly users.

Still, the framing the authors land on is the useful one to carry forward: migration as a multi-homing strategy that insures against governance uncertainty, with stickiness on the new platform decided in the first few reconnections.