Research: crowdsourced fact-checking can improve social media moderation by channeling partisan motivations to be productive if opposing sides police each other
and most experts — see #misinformation as a “major problem” and believe internet companies should do more to curb its spread. - via @FT By Prof. @DG_Rand @MIT #Tech https://www.ft.com/... David G. Rand / @dg_rand : have a piece in @FT today on work w @_JenAllen @Cameron_Martel_ examining @Twitter @birdwatch approach to using crowds to identify misinformation, which we think is very promising bc of - rather than in spite of - partisan motives https://www.ft.com/... Kevin Loker / @kevinloker : “channelling partisan motivations to be productive, rather than destructive, is critical for platforms that want to use crowdsourced moderation.” https://twitter.com/... Jennifer Allen / @_jenallen : 🚨Paper alert!🚨Birds of a Feather Don't Fact-check Each Other #CHI2022 On Twitters crowdsourced factcheck program @birdwatch, partisanship is king: users mostly flag counterpartisans' tweets as misleading & rate their notes as unhelpful https://psyarxiv.com/57e3q 1/ https://twitter.com/... @1br0wn : “engaging users in...moderation...can help platforms satisfy critics from both sides — identifying #misinformation at scale, while avoiding claims of top-down bias. Thus, there is a strong business case for platforms investing in crowdsourced fact-checking.” https://www.ft.com/...
Context & Ripple Effects
Twitter's Birdwatch experiment has traveled a rough arc: an early analysis of the pilot found partisan rhetoric and missing source citations among its roughly 1,000 users, and by March 2022 it remained a small pilot with only 359 active contributors. The CHI2022 paper from Jennifer Allen, David Rand, and Cameron Martel explains why it worked anyway — activity was driven by partisanship, with users flagging counterpartisans and rating their notes unhelpful.
The FT piece argues this is Birdwatch's feature, not its bug: cross-partisan policing filters out one-sided notes, which is why the mechanism later matured into Community Notes, a surprise hit built on crowd-sourced consensus. That matters now because platforms are pulling back from professional fact-checking partnerships even as false AI content spreads.
First-order effects
- Platforms running consensus-based systems like Birdwatch get a design rule for free: don't suppress partisan motivation, structure it so each side rates the other's notes, since agreement across hostile camps is the credibility signal.
- Researchers studying moderation gain a validated alternative to top-down labeling — one that addresses the unpaid-moderator problem platforms have offloaded onto journalists and volunteers since at least 2018.
Second-order effects
- Rival platforms face pressure to adopt cross-partisan rating mechanisms rather than expert or in-house review, because Community Notes' perceived fairness becomes a competitive differentiator on trust.
- Professional fact-checking organizations lose leverage in platform partnerships just as the number of fact-checking sites worldwide declines — crowdsourced consensus is eating the market they were built to serve.
Third-order effects
- If adversarial cross-rating holds up at scale, platform governance shifts from gatekeeping (who decides what's true) to mechanism design (how rival incentives cancel out), making moderation auditable through the rating data itself.
- The same structure becomes a candidate defense layer against synthetic media, where human consensus filtering may be the only scalable check on machine-generated falsehoods.
The trend: Content moderation is moving from centralized expert judgment toward adversarial crowd-consensus mechanisms, with partisan bias harnessed rather than eliminated.