Fact Checking Biases

From iGeek
Jump to: navigation, search

There are many forms of bias when doing studies or trying to analyze data. When Fact-Checking the Fact-Checkers, or the filters we put on what's real, it helps to keep a few of those in mind, like: #Selection Bias (what they pick), #Oversampling Bias (picking the same thing over, or minute variants), #Standards Bias (not holding both sides to the same standard), and so on. If you can show that a site has any of these biases, then you discredit their claims of impartiality or objectivity. If you can show they do all of these, then you wreck their credibility... at least for those who aren't biased themselves. Anyone that refutes that this evidence matters, proves that they are not to be taken seriously (as objective).

There have been a few studies done, on one or another Fact Checking site, that proves that kind of bias for that site -- yet, you can and should probably genericize the results to all Fact Checkers; the problems of bias in journalism are systemic (don't seem localized).

These problems are lightly concerning for trusting the validity of a single fact check, but as any single error is localized, it's not a big deal. But New Media (the Internet: Google, Facebook, the flock) is starting to use their own echo-chamber of fact-checkers to validate what is real or FakeNews in general. That is far more concerning -- as they scale their flawed solution, they're magnifying the problem. And they miss the fatal flaw: if all the samples have the same problem, then adding more samples doesn't fix it. This is how the aggregate polls were wrong in the 2016 election, because all the individual polls were wrong in the same way (read Trump Election: Orange is the new black. GIGO: Garbage-In = Garbage Out. Some rubes are doing that with our "news" or putting that herd filter their reality. The Germans, Italians and Japanese did that just prior to WWII, and it didn't work out too well, for the same reasons.

Selection bias

🗒️ NOTE:
If you look at actually psychological honesty studies by party, at best these numbers would be loosely even, and at worst, they would skew the other way. The reason isn't that the Republicans are more honest, it's just that the majority of the Press being biased left, which means they hold Republicans to a higher standard and Democrats to a lower one. So by nature of incentives and consequences, Republicans have to be better or the Press will hammer them, thus they are.

PolitiFacts worst sin is in how they pick what to review (selection bias). George Mason University Center for Media and Public Affairs did a study, and even the NYT ran a piece on it (written by PolitiFact editors) which showed "Republicans lied 3x as often as Democrats" (especially ones PolitiFact doesn't like). [1] The leftists and their media use that to show, "see, Republicans are bigger liars". But the facts are there's no evidence that the Republicans lie any more than Democrats.

What the study really showed was the PolitiFact was 3x as likely to pick a Republican lie to complain about than a Democrat one. (And/or hold them to a different standard). Even during years where Obama is embroiled in lying about Benghazi, the "fact checkers" were calling Romney pants-on-fire for doubting the President's claim that this was a spontaneous attack by people mad over a video that no one heard of, and they just happened to be carrying mortars and RPG's with them. Instead of admitting Obama's claim false, the Fact checkers called Romney's accusations that Obama lied, unsupported by the evidence at the time -- and never corrected the record later when later evidence came out that proved that Romney was correct and Obama had been lying all along

So from the same study, the Democrats heard what they wanted to, that Republican Politicians lie more, and they repeat the mythos. And Republicans and Independents saw a glimpse of the truth, that partisan democrat cheerleaders posing as fact checkers are far more likely to use their jobs to attack Republicans and make excuses for Democrats than to be unbiased purveyors of truth.

Standards bias

The simplest example of standards bias is the following example of Politifact's anti-Republican bias:


But there are many examples like that.

MSNBC's Ed Schultz makes an exaggerated claim that's off by at least 33% (and more like 50%) -- and it's "half true". Conservative talk show host Larry Elders is off by less than 25%, and he's "Mostly False".

When Ted Cruz cited a well researched and well known claim that "most violent criminals are democrats"[2]. PolitiFact lost their nut, and picked nits about how:

  • the study cited, "didn't differentiate between violent criminals and all criminals", as if that's going to really change the outcome.
  • they whined about correlation is not causation, but Cruz's point wasn't that voting democrat will make you a felon, his point was that the reason the DNC is soft on crime, is because violent felons are more of their base. (They are).
  • Then PolitiFact didn't research any of the dozens of other papers, as there have been many, in many states about how felons vote, or what racial/cultural and socio-economic demographic they are (and how those demographics vote)

The result? They declared something that was completely true, was "mostly false". Hey, it's good enough for the Democrats (their readers).



  • For Trump's top 5 lies, like Large Scale Voter Fraud, Not supporting Iraq War, or Obama's birth certificate, they not only use selection bias on what they check (or choosing how it was phrased to make it look worse), and judge it by a harsher standard than Hillary, but they check the same lie 5, 6 and 7 times each.
  • For bigger lies by Hillary like "was named after Sir Edmund Hillary", "Landed Under Sniper fire in Bosnia", or "Did not send classified info in emails", they checked the same lie, 0, 0 and 2 times (the latter having one true and one false, which brings into question their ability to understand what facts are).

Then there's how many times you get checked at all (a form of selection bias) -- for example Marco Rubio was fact-checked as many times before he took office than long-time senator and erstwhile Senate Majority Leader Harry Reid has been checked in his entire career. Ted Cruz was checked more than any Democrat Senator, and Paul Ryan was fact-checked nearly 2:1 as often as House Speaker Nancy Pelosi was. Michelle Bachmann was checked about as much as vice presidential candidate Tim Kaine.


Other than cherry picking what they "FactCheck", and oversampling Replicants and under-sampling or ignoring Democrats, and having completely different standards for what's Mostly True or False, and other than distorting the questions and answers to make Republicans look worse and Democrats look better, the FactChecking sites are completely fair. But that's like saying other than being a serial rapist, he was nice towards Women.


📚 References