Misinformation & Community Norms

The dominant framing of misinformation treats it as a problem of individual judgement: people believe false things, so correct them. This project argues that framing is too narrow to explain what actually happens in online communities, and too narrow to design for.

What the surrounding community does turns out to matter as much as the content itself. The prevalence of a behaviour in a community shapes how individuals respond to misinformation in it — and individuals' own responses can, in turn, blunt the effect of that prevalence. Interventions built only around the individual reader miss this entirely.

The project also holds the lab's cross-platform work on how discursive framings of a crisis diverge between public, media, and government discourse, and on the downstream relationship between divergent narrative exposure and conflict disposition.

Synthetic media belongs here rather than in a category of its own. Machine detection of AI-generated imagery has improved faster than anyone's ability to make that detection useful to the person actually looking at the picture: a classifier that is right 98% of the time in a benchmark does very little for someone scrolling past a face they had no particular reason to doubt. The lab works on both halves — how deepfakes are created and detected, and then the harder part, which is what an end user can actually do with a detection signal. Where a warning belongs in an interface, what it has to say to be believed, what it costs when it is wrong, and who is worse off when it is absent. That question descends directly from DART, which built deception resilience by rehearsal rather than by warning label.