Social media platforms made progress since the 2016 election in addressing the most blatant abuses, but the misinfo issues may be endemic to their core product
Will Oremus / OneZero : Tweets: @ghoshd7 , @dlberes , @willoremus , and @whotargetsme . Thanks: @dlberes Tweets: Dipayan Ghosh / @ghoshd7 : “Four years on, it's hard to substantiate a claim that the big tech companies are doing any better than they did in 2016.” Spoke with @WillOremus whose excellent column for @ozm should raise concern about social media ahead of Election Day. https://onezero.medium.com/... Damon Beres / @dlberes : Don't miss this great rundown from @WillOremus about how social media companies are positioned to deal with... <gestures wildly> everything... leading into Tuesday https://onezero.medium.com/... Will Oremus / @willoremus : In the end, the hidebound @dlberes managed to censor this sentence down from its original 12 nouns to a positively repressive 10. https://rsci.app.link/... https://twitter.com/... Who Targets Me / @whotargetsme : Balanced summary of what's happened on social media since the 2016 election from @WillOremus. Some good, still a lot of bad, largely caused by the structure and design of social media itself. https://onezero.medium.com/... Thanks: @dlberes
Context & Ripple Effects
Will Oremus' pre-Election Day column lands at the end of a four-year arc the corpus has tracked closely: the [[a:935379|demonstrable progress Facebook and others made against misinformation during the 2018 midterms]], followed by [[a:935279|analysis of 250K posts and 5K political ads showing how much politically motivated misinfo still slipped through]]. The throughline is uncomfortable for the platforms — Dipayan Ghosh's quoted verdict is that it is hard to substantiate any claim they are doing better than they did in 2016.
What changed by late 2020 is less the volume of abuse than its location: the problem increasingly looks like a property of the product itself rather than of bad actors exploiting it. Oremus had already argued in August that Facebook, Google, and Twitter were repeating the news industry's old mistake of placating hyperpartisan critics who are 'working the refs' — and his election-eve framing shifts the question from enforcement failures to engagement-based design.
First-order effects
- Facebook and Twitter enter Election Day leaning on warning labels and takedowns of the most blatant abuses — measures aimed at the visible symptoms while leaving the recommendation engines that surface misinfo untouched.
- Dipayan Ghosh and other former insiders publicly refuse the platforms their preferred narrative, making 'we've improved since 2016' untenable as a PR line heading into the vote.
Second-order effects
- Enforcement that works just moves the bottleneck: YouTube's cuts to fringe-channel recommendations ended up making Fox News the most recommended source for election videos, showing that dialing down one supply of misinfo reroutes demand to mainstream partisan outlets.
- Warning labels delivered a public-relations win but arrived so late in the information cycle that they functioned as PR more than as product — pressuring rivals to match the optics of labeling without changing underlying ranking incentives.
Third-order effects
- If misinfo is endemic to engagement-optimized feeds, no amount of policy-team iteration fixes it — pointing toward structural answers like algorithmic accountability regulation or design mandates rather than content-by-content moderation.
- The trajectory holds beyond 2020: by 2023 Meta and YouTube, battered by the moderation wars, were following X in retreating from policing political misinformation altogether — suggesting the 2016-to-2020 cleanup was a peak of effort, not a floor.
The trend: Platform misinformation policy is cycling from post-2016 remediation toward abandonment of political-content policing, because engagement-based design keeps regenerating what moderation removes.