The majority of Americans were not exposed to fake news during the 2016 election

The argument

The claim that social media was used to target disenfranchised voters in swing states presupposes broad reach: a campaign of misinformation can only move an electorate if it actually arrives in front of a large and relevant share of it. Research on the 2016 election casts doubt on that premise. A study of more than 16,000 Twitter users found that only a small fraction of the content circulating during the presidential election was fake news, and that its distribution was highly concentrated — a small number of users saw the majority of the misinformation, and those users were typically older and politically conservative. Both halves of that finding matter. The first concerns volume: if fabricated content made up only a small share of what people encountered, it was competing against a far larger body of ordinary political material rather than dominating anyone's information diet. The second concerns distribution, and it is the more damaging to the targeting claim. Misinformation reaching overwhelmingly the same narrow group — older, already conservative, and by that description already committed — is not reaching the disenfranchised or the persuadable. It is landing on people whose votes were least in doubt. A targeting operation of the kind alleged would produce a different pattern. One would expect exposure spread across the marginal and the disengaged in contested states, not concentrated in a demographic slice defined by age and existing partisanship. The evidence shows the second pattern, not the first. The conclusion proponents draw is that the effect of fake news, and particularly its effect in swing states, may have been considerably less pervasive on social media than was originally thought — and that social media was not, in the manner alleged, used to target the disenfranchised.

Premises

Counter-arguments

Even a small amount of fake news could have an impact on the political process and therefore on the US presidential election.

Rejecting the premises

[Rejecting P1] One platform's users are not the electorate, and a study of this kind measures sharing of content from a defined list of sources. It cannot see material circulated through private messaging, closed groups or paid advertising — which is exactly where targeting of the alleged kind would be expected to happen. [Rejecting P2] Narrow reach is compatible with decisive effect. Targeting means concentration by design, so a small, specific audience in a handful of states is what the opposing position describes rather than evidence against it — and one election is a thin basis for a claim about a continuing practice.