Why social context matters when correcting vaccine misinformation
In a paper published this year in Vaccine, I asked a question that has become increasingly urgent: can ordinary social media users — not platforms, not experts — effectively correct vaccine misinformation? And does it matter who is doing the correcting?
The set-up
We ran a pre-registered online experiment on Prolific using a hypothetical "COVID-66" pandemic scenario. Participants were exposed to anti-vaccine social media posts and then saw one of several types of correction: a platform-generated label, an expert refutation, or a "social correction" from another social media user. We measured both their factual beliefs about the vaccine and their vaccination intentions.
What we found
Social corrections worked — but in a way that surprised us. The correction's effect on intentions was mediated less by belief change than by changes in perceived social norms. When a correction came from a peer, participants updated their sense of what other people thought about vaccination, and that shift predicted intention change. Belief change, by contrast, was a weaker pathway.
The lesson: misinformation is not only an informational problem. It is also a normative problem — it tells people what others think.
Why this matters
Most platform interventions today are designed to correct facts. They add labels, link to expert sources, or suppress content. These are useful, but they may miss half of what misinformation actually does — which is to distort our sense of what is socially normal.
If social norms are part of the mechanism, then social corrections — visible, peer-generated counter-messages — should be part of the response. This has implications for how we design social media interventions, and how we think about the limits of fact-checking.
What I'm working on next
I have a follow-up paper under review at Scientific Reports that directly compares algorithmic refutations (the kind platforms typically deploy) with social refutations. The early evidence: when correction comes without social context, it does much less work. More on that when the paper is out.