In case you hadn’t noticed, I am really sick and tired of writing about l’affaire Reeves.
Needless to say, both campaigns seem heavily invested in the entire affair, with Team Reeves investing in a time clock (that is getting panned fairly negatively) and an op-ed while Team Vogel not only doubling down with an op-ed but sponsoring content on Virginia Right of all places.
Don’t… just don’t… click on any of the links. Your time is valuable, my time is valuable. Needless to say, folks are talking about it.
Here’s my point in all of this, and it has been a salient point from the beginning that unfortunately has cost me a friendship in the Virginia Senate — if for no other reason than I simply didn’t want to go 100% with 80% of the facts.
Let’s assume for a moment that Reeves really does have most of the facts — a generous 80% chance that Reeves information and conclusion are sound. That seems fair, yes?
Now let’s introduce the great variable in all of this — the defense that Vogel is the victim of wardriving. Is it plausible? Is it ridiculous? Let’s say that — given our 80% confidence interval with Reeves’ claims that the counterpossibility of wardriving is a generous 1 in 10 chance. Let’s further stipulate that if — on the wild chance Reeves is wrong on this — that the odds that Vogel was indeed the victim of wardriving flip entirely to a 80% confidence interval (what other explanation could there be, plus this would be the same credence we would offer to Reeves current claims — so it seems like a fair number iff Reeves narrative is disproven).
Stay with me here. If P(A) of Reeves’ Claims is 80%, and P(B) of Vogel as Victim is 10%, and further that if Reeves’ claims were to be disproven that the scenario that Vogel was the victim goes to 80% — the given what we know today… what are the odds that Vogel is the victim of a wardriving incident?
1 in 3.
Welcome to Bayesian Modelling, folks.
I’ll give you an example. Let’s say that 5% of the human population your age and with your hereditary background have cancer. Let’s further say that the blood test for cancer is 80% effective, and 20% of the time will give you a false positive. You take the test… and it comes up as positive. What are the odds that you actually have cancer? Not 80%… not even 20%… but 17.4% — meaning that there’s a 5:1 probability you don’t have cancer even if the test is right 80% of the time.
Just to show you how this math works in l’affaire Reeves, click here for our data assumptions. Play with the math all you want… it may change the margins, but it really doesn’t effect two salient points we’re trying to demonstrate here: (1) neither party has a 100% lock on the chain of events, and (2) the principle point for resisting Reeves’ narrative to date that Vogel is out-and-out wrong is solely because the confidence interval doesn’t add up.
Now here’s something that should comfort Team Reeves in all of this, namely that the confidence intervals work both ways on this. For instance, the probability that what we know of P(A) being true and ¬P(B)? 94.74% — meaning that the information we are seeing thus far is probably going to be accurate.
…but it’s not ironclad. Which has to be frustrating for those who instinctively and even intuitively look at the claims and say there’s no possible way that Vogel is innocent.
Now the smart reader will note that 94.7% plus 33.3% don’t add up to 100% and assume this is a problem. Don’t think of them as numbers… but rather as a cloud of information points. The confidence interval that Reeves’ story to date is accurate is 94.7% spot on. The confidence interval that Vogel was the victim of wardriving is 33.3% spot on. That haze (if you will) between the two is the ambiguity that makes the claim “Vogel did it” impossible at this stage.
Hence the reason why Reeves can’t stay at 80% and needs to get to 100%. Reeves himself knows that he doesn’t have all the information, and to his credit admits as such. There’s one step that’s missing, and Reeves is demanding that the information be released — information that Vogel isn’t bound to do (yet), of course… but it makes great political theater asking for it. Nevertheless, not having 100% confidence in the claim by having only 80% of the information creates that gap of the unknown, that uncertainty.
That’s why the probability of P(A) and P(B) being the case is so high — that being a 33% chance that Vogel was the victim of wardriving.
In fact, such a gap of uncertainty means that — not being certain of the claim of impropriety himself, Reeves has committed the very error he accused Team Vogel of doing — spreading false information based a probability, but not the truth. The intuitive focus on the 94.7% was right… but it ignored the 33.3% on the other side.
Now naturally, one could come back and argue that this presents a 66% chance that Vogel (or the Vogel’s… or Team Vogel… pick your poison) was the source. Small problem with that as well, because the probability of a case doesn’t mean one gets to make the ironclad allegation that it did happen with absolute confidence.
In other words, one shouldn’t ethically go to the Washington Post with the prospect 33% doubt, even if you have 94.7% confidence in 80% of the case. Nor do you demand that the candidate bow out of a statewide race being 33% unsure of one’s case — especially as thus far, the case is entirely one sided with Vogel having yet to defend herself in as rigorous a way as Reeves has defended himself.
Unfortunately, we now have two camps making competing claims based on assumptions that while seemingly factual are not truth. If either Reeves or Vogel get their confidence interval up to 100%? The problem resolves itself, naturally. But so long as there is lingering doubt? One can see that even a distant probability can open a very wide door that should give the adults in the room pause.
Bayesian Modelling demonstrates a good lesson when it comes to negative campaigning, allegations of scandal, rumormongering, and the like. Odds are if you think through it, our facts may feed a narrative (fake news?) and may even give us confidence in what we see and believe, but rarely does this confidence interval afford us a window on lock-solid truth.
In fact, when you get the hypothesis wrong? You blow elections… just like Hillary Clinton did in believing she was up in Wisconsin, Michigan, and Pennsylvania — all done on Bayesian modelling, folks.
(Helps to have done a bit of statistical modelling when you worked for the U.S. Drug Enforcement Administration as well, so long as we are measuring out resumes.)
…and that’s why we don’t lead with 80% of the facts, even at the deeply regrettable cost of a long-standing friendship who believed with 94.7% confidence in the 80% of the incident his surrogates offered — precisely because there’s a 33% chance that getting from 80% to 100% just isn’t going to pan out the way you want it to. That’s runs against the grain of everything I personally believe… and even at the cost of perceived loyalty and ignored counsel to friends, there is no honor in winging another person based on innuendo and assumptions — no matter how grounded we may think those assumptions may be.
My personal ethics and background (hard to explain as they are) just can’t go for that ride.
Slander takes many forms. Without the lock-solid truth, we need to be far more careful about absolutist claims — especially among those we call friends.
The only 100% we can be certain of at this rate is that none of this sordid affair smiles kindly on what we believe as either conservatives or Virginians. Surely we can do better… surely given the tremendous issues we are facing in Virginia, we can find someone willing to run on issues and ideas rather than who-shot-John?
A forlorn hope, perhaps. I’m not alone in it.