Saturday, March 18, 2017

Facts, inferences, and journalism when discussing Trump's bullshit

Nonpartisan journalists operate on the model that they limit their reports, to the degree possible, to what they think are verifiable facts.  How do they verify facts?  Documents, video evidence, etc.  So, when Trump claimed that Obama wiretapped him, in the absence of contrary evidence, most nonpartisan journalists were unwilling to simply call bullshit.  Why?  Because in theory, there could be evidence somewhere.*  Hence, the claim could be true, Trump could have evidence, and he might not be a total fucking liar.  I'd tell you to stop laughing, but no, go ahead and laugh.  Anyone who ever took this seriously was a fucking moron, but that has always been true for Trump's lies, which I have analogized to the Nigerian Prince scam (see here and here).**

The problem with applying the verifiable-facts-only model in the case of the wiretapping claim is that we know exactly how it came about.  Radio wacko Mark Levin first came up with it, and from there, it spread to... Breitbart!  From there, it was a straight shot to Trump himself.  We also know that Trump does not take regular intelligence briefings because "I'm, like, a really smart person," and that he hates the intelligence agencies because they tell him things like, oh, say, Russia hacked the DNC in order to interfere in the election to put him in office.  That's why, when pressed for his sources on the claim, he cites, not briefings from intelligence agencies, but Fox News personalities who are dutifully repeating what Trump says because he says it in a perfect circle of terrifying, paranoid lies.

So, do we evaluate the claim based on what we know about the process by which the "information" is transmitted, giving it an essentially zero probability of being true, and thereby treat it as a total fucking lie, or do we treat it as a question mark because it is an unverifiable claim with no evidence?  The facts-only journalistic model says the latter.  My statistician/social scientist brain, updated in Bayesian terms based on Trump's record of lying, giving him zero credibility, says the former.

This is an important point about the gap between how journalists think about evidence and how statistically-minded social scientists assess claims.  I keep thinking in statistical terms.  Yesterday, I wrote about Bayesian assessments of Trump.  Journalists don't, and can't think in those terms, partially because they are conditioned to think in more legalistic terms (you can't convict on probabilities), and partially because, well, journalists generally can't think in mathematical terms anyway.  However, I will always revert to Occam's Razor.  The simplest explanation, when Trump makes a wild claim with no evidence, is that he is full of shit.  Standard journalistic models don't work when the President is a pathological liar.

*  In theory, communism works.  In theory.-- Simpson, H.J.

**  I'll admit that I dragged that series out to stall for time before politics got going again, but there's some good stuff in that series.


  1. I'm still chewing on something you said last week that resonated:

    p(A) vs p(A|Trump said A). I'm also tossing in there: p(A|Trump said B).

    Still not sure where I'm coming down on it. Is Trump a liar or a bullshitter? What if he's both? How to I update p(A) after Trump speaks?

    It's a fascinating problem, I think, once you mix in that Trump could be strategically choosing between lying and bullshitting.

    1. And then there's the problem that some of what he says, he actually BELIEVES. A bunch of those anonymous sources claim that he really believes the wiretap stuff. So, you've got bullshitting, lying, and delusions. How complicated do you want to make this?

    2. Well, you've heard about my attempts to come up with a model for veto threats that incorporates bluffing, appealing to audiences, past behavior, and preexisting opinions about truthfulness.........

      ...if you think it's possible to make a model of Trump's behavior tractable, I'm all ears. My sense is that equation has WAY too many fucking variables to ever yield anything more than statements about the numerical relationship between Trump's ego function and the GOP's discount factor.
      (aka, sophistry).

    3. But, yeah, I hadn't factored in "batshit crazy" into my thinking yet.

      Model that shit.

    4. I believe someone did. His name was Schelling.

    5. Crazy like a fox is not batshit crazy.