How Facebook and Twitter fight online voter suppression for the midterms with algorithm changes and partnerships with organizations that report disinformation
and they're already seeing fake accounts, ppl sharing wrong election dates, false reports of ICE agents at polling places and more http://www.washingtonpost.com/ ... @wired : Misinformation now spreads farther, faster, and ensnares unwitting accomplices who share bad information without realizing it. Here are a few general themes that come up every election cycle. http://www.wired.com/... @cnetnews : The automated accounts were quickly caught, Twitter said. http://www.cnet.com/... Peter Skomoroch / @peteskomoroch : This isn't really a win for Twitter. It's a black eye that shows they had to react to reports from 3rd parties to detect these fake accounts, rather than detect them internally with machine learning. http://twitter.com/... Nick Pacilio / @nickpacilio : - 10M Tweets already about voting/early voting - Already more #Midterms2018 Tweets than '14 - 1,000 candidates have a Twitter election label - 15,000 US users have ‘vote’ in their display name - #IVotedEarly, #yovoté, #IVoted digital ‘i voted’ stickers http://blog.twitter.com/... Binyamin Appelbaum / @bcappelbaum : The key words here are “after the party flagged the misleading tweets.” http://twitter.com/... Tony Romm / @tonyromm : NEW: Twitter and Facebook are trying to ramp up their efforts to combat voter suppression online. Both have already witnessed a flood of posts aiming to deter immigrants with citizenship from voting. My Friday longread: http://www.washingtonpost.com/ ... Thanks: @tonyromm
Context & Ripple Effects
Days before the 2018 midterms, Facebook and Twitter are running their most visible election-integrity operation yet: tweaking ranking algorithms to demote suppression content and outsourcing detection to partner organizations that flag disinformation. The trigger is a live wave of suppression posts — wrong election dates, false reports of ICE agents at polling places, and what the companies describe as a flood of posts aimed at deterring naturalized-immigrant voters.
The playbook was already showing strain in real time: Twitter says its automated accounts were caught quickly, but commentators like Peter Skomoroch read it as a black eye rather than a win, since the company reacted only after third parties flagged the tweets. The verdict came fast — Slate's post-midterm assessment found demonstrable progress on misinformation but argued engagement-based algorithms keep the problem alive, while Jonathan Albright's analysis of 250K posts and 5K political ads documented how much politically motivated misinformation still slipped through.
First-order effects
- Twitter and Facebook are directly suppressing a specific genre of content — fake election dates, fabricated ICE-at-polling-place reports, and posts targeting naturalized citizens — with Twitter dependent on external reports rather than proactive detection for misleading tweets.
- Partner fact-checking organizations gain operational roles in US elections, effectively becoming outsourced moderation infrastructure whose flags determine what the platforms act on.
Second-order effects
- Because takedowns sit downstream of engagement-ranked feeds, the platforms' own distribution systems keep amplifying the very misinformation their teams remove — forcing the next iteration of the playbook, which reappears nearly intact for the 2019 Democratic debates and again for Election Day 2020.
- Detection burden shifting to third-party reporters sets up a competitive asymmetry: whichever platform builds faster internal detection needs fewer partners, turning moderation capability into a differentiator between Facebook and Twitter.
Third-order effects
- If the pattern holds, election-integrity response hardens into a recurring operational cycle — each US election triggering algorithm changes and partnership activations — rather than a one-off fix, with the underlying engagement-based recommendation architecture left untouched.
- Recommendation systems themselves become the contested layer: the trajectory runs from removing bad actors to tuning what gets recommended at all, a shift YouTube's 2020 experiment made concrete when cutting fringe-channel recommendations simply redirected election-video traffic toward mainstream outlets like Fox News.
The trend: US platforms are institutionalizing election-cycle disinformation response as a repeatable algorithm-plus-partnership playbook, while leaving the engagement-driven recommendation systems that spread misinformation structurally intact.