The Trouble With 'Half True' Fact-Checking Verdicts
Under ideal circumstances, fact-checking outlets offer simple ratings that inform readers whether a claim they have encountered is true or false. As with many situations in life, however, reality is often far more complicated than the ideal. Regular readers of fact-checking sites might encounter clear verdicts such as “True” and “False,” but they are also highly likely to encounter more nuanced verdicts such as “Mostly True” or “Mostly False.”
Perhaps the most complex verdicts employed by fact checkers are “Mixture,” which is frequently used by Snopes, and “Half True,” a favorite of PolitiFact. Just how frequently are these verdicts used? A search of the last 120 days of data entered into the RealClearPolitics Fact Check Review showed that Snopes used the “Mixture” verdict in 44 out of the 266 fact checks we examined during this period, or almost 17 percent of the time. PolitiFact used its “Half True” verdict in 78 out of 378 fact checks, or about 21 percent of the time. In short, they use these verdicts a lot.
As I pointed out in a column last week, there are times when such inconclusive verdicts, though imperfect, make the most sense. In last week’s example, Dan MacGuill of Snopes pinned the “Mixture” verdict to an article that, while substantively true — it “faithfully reported” the facts, in MacGuill’s words — carried a false headline. It would be unfair to call the article in question simply true, because the headline was inaccurate. But on the other hand, it seemed unfair to label a faithfully reported article as false, when the headline was the sole untruth.
But the “Mixture” verdict can also complicate matters. For instance, MacGuill wrote a piece for Snopes assessing the claim that singer Pharrell Williams sent President Trump a cease-and-desist letter for using his song “Happy” prior to a speech at an Indiana convention. Williams’ attorneys did send such a letter to the president. However, through his reporting, MacGuill found that it was the organizers of the convention, and not the president or his staff, who made the music selection. This additional information is certainly important, and it is a credit to MacGuill’s diligence to have included it. Yet the claim that MacGuill was investigating — that Trump was sent a cease-and-desist letter — is true.
A recent PolitiFact piece by Stephen Koff checked a claim by Rep. Troy Balderson (pictured). Balderson, a former Ohio state senator, asserted that as a result of the Affordable Care Act, “there are counties out there, in my former Senate district, that didn’t have doctors and providers.” Koff assigned a “Half True” rating to Balderson’s claim. He wrote that “[n]early half of all Ohio counties … wound up with only a single ACA carrier this year. For some people, that changed their access to the doctors and hospitals they preferred.”
By this account, Balderson’s claim seems largely true. But, added Koff, “things will improve for many in 2019. And even counting all the changes since the ACA began, many more people have health coverage now than before.” Koff provided the rationale for his verdict, writing: “If you are a constituent who lost access to your preferred insurer or provider, you might have called Balderson to urge him to fight for change. If you are a citizen who got health care because of the ACA, you might have joined a protest in hopes Congress will leave it alone.”
In other words, Balderson’s statement is true, but that’s a wash because other people might have had different experiences from the one he described. The “Half True” verdict doesn’t offer much in the way of clarification in this case. It essentially puts the call back into the hands of the reader, and an ostensible inquiry into fact or fiction is rendered an exercise of opinion.
While indeterminate verdicts might present problems of clarity for the reader, there is an even more troubling wrinkle: censorship. Facebook has confirmed for us that a “Mixture” rating from a fact-checking organization is enough to issue a warning to Facebook users who try to share stories assigned such a rating. If users share the stories anyway, their “distribution” will be “reduced,” meaning the stories will be made less visible in the feeds of other Facebook users who might happen upon the shared post. We have inquired with Facebook as to whether stories assigned the “Half True” verdict are treated the same way and have not heard back — but since “Mixture” and “Half True” are roughly equivalent, it stands to reason that this is the case.
In this age of disinformation, readers look to fact checkers for clarity -- distinct calls as to truth and untruth. But often, due to the complexity of the real world and judgments made by the fact checkers themselves, such clarity is not provided. Perhaps the fact-checking outlets would better serve their readership if they added a rating of “True*,” with the asterisk indicating that while the claim in question is verifiably accurate, there is more that should be considered by the reader. But until such a reform is made, this is yet another indication that readers interested in truth need to look past the headlines and evaluate all available information.