The Limits of Reason
I’ve withheld comment on the George Zimmerman trial because it’s not an edifying subject—it’s intellectual junk food, a soap opera. I agree with James Goulding that our blogs should elevate the discussion, which, if we want Dark Enlightenment ideas to spread, is a better strategy than writing screeds about “the Jews control America!” or “fuckin niggers at it again!”
However, the latest round of Zimmerman punditry deserves comment.
Matthew Yglesias, who is, if possible, an even whiter Spaniard than I am, writes the following in a post entitled “Bayes’ Theorem for Dummies”:
I think what Cohen really means to be arguing isn’t so much that neither he nor Zimmerman are racists, but that racism is the correct social and political posture. That white people have good reason to fear black men, and that therefore all black men should be put in a subordinate position. But as a logical argument, Cohen here is falling afoul of very poor statistical inference . . .
. . . the fact that young black men are disproportionately likely to be involved in violent crime in no way licenses the inference that you should stop random black men on the street and begin treating them like criminals.
For example, since moving to a majority black city 10 years ago, it is the case that 100 percent of the people who randomly assaulted me on the street were African-American. And yet that was a single incident on one day out of thousands. The overwhelming preponderance of black men I walk past on the street on a day-to-day basis—even the young ones, even the ones wearing hoodies—aren’t committing any violent crimes. If I were to start questioning every single black male teenager I come across as a criminal suspect, I would very much be engaged in unreasonable behavior.
The critique is elementary. Formally, we may say that Zimmerman confused Pr(A|B) with Pr(B|A). Colloquially, we can summarize Yglesias’s critique thusly: “Even if most criminals are black, it doesn’t follow that most blacks are criminals. So it’s simply not logical to profile every black person you meet as being a criminal threat.”
Yglesias probably thinks his point is unanswerable. I’ll give him credit for at least attempting to return an element of logos to the seething pathos of this turgid affair. However, like the young college girl who first discovers that American settlers didn’t treat the natives too kindly, Yglesias is satisfied with setting his feet into the cement of his first rational step and not moving anywhere else.
From my perspective, his point is entirely answerable, and on purely rational grounds. It’s all a matter of returning logic to its context and filling in the priors.
First, Zimmerman was not following “all black men,” as Yglesias implies later in his post. Zimmerman was following a singular young black male, dressed in ill-fitting clothes, wandering a suburban neighborhood at night after a recent streak of burglaries committed by young black males in his neighborhood. 100% of recent burglaries had been committed by a marked population that constituted only a small percentage of the community’s whole population. Trayvon Martin, at various points of description, fit the marked “profile” of recent burglars.
As commenter Cail Corishev writes at Sailer’s:
I’m writing a Bayesian spam filter (based on Paul Graham’s Plan for Spam), and “profiling” is exactly what it does. There’s no one word that guarantees that a message is spam. But if a message contains 10 words that appear frequently in spam, and it doesn’t contain any words that appear exclusively in non-spam, the probability that the message is spam will be very close to 1.
That’s what profiling means. It doesn’t mean, “Stop all blacks because blacks are more dangerous than other groups”; that’s what liberals like Bloomberg do because they’re trying to avoid profiling. Profiling would have a cop say (pulling percentages out of my hat for the example), “Ok, there’s a young black man walking down the street in this neighborhood, so historical data says there’s a 5% chance that he’s up to no good. That’s not nearly enough to suspect anything. But he’s also wearing a hoodie, which adds another 5% (whether he’s black or white), and he’s hiding his face (another 3%), and keeping his hands shoved down deep in his pockets (another 10%), and his sneakers look brand new, which we’ve been told to look out for because a store was knocked over last night (another 20%). Let’s pull over and ask him where he’s headed …. Okay, he doesn’t seem to know this neighborhood (another 20%) so let’s chat with him a bit more…. Ok, he showed us what he was holding in his pockets, and it was liniment and denture cream that he said he’s taking to his grandma whom he’s staying with a few blocks from here (-30%), and he seemed friendly and relaxed while we talked to him (-10%), and offered to show us the receipt for his shoes (-30%). Seems okay, tell him to have a nice day.”
My limited knowledge of Bayesian statistics is, indeed, that it provides a framework for working with multiple priors for coming up with an averaged statistical probability. In other words, its entire point is to move away from the kind of decontextualized logic that Yglesias invokes, a logic that would make sense as a critique if Zimmerman (or anyone) had indeed been following every black male he met on a daily basis. But no one does that, so Yglesias’s invoked logic is really just a shot at a strawman.
Second, until the fight itself, Zimmerman’s actions were quite modest. The act of “profiling” or “being suspicious” is an act of personal caution, not a willful attack on someone else’s personal rights. (My crossing to the other side of the road while being approached by a group of young black males does not in any meaningful way affect the young black males. Zimmerman’s following Trayvon did not affect Trayvon.) Without his strawman (stop every black on the street!), Yglesias, I imagine, might modestly modify his claim: Zimmerman’s decision to follow Trayvon, or my decision to cross the street, remains illogical because, well, the same old same: “just because criminals are often young black males, most young black males aren’t criminals!”
The comments on Steve’s post provide many obvious rejoinders, which I paraphrase here:
The vast majority of grizzly bears don’t attack hikers, so it’s irrational for hikers to carry bear spray in the wilderness!
80% of lumps aren’t cancerous, so my mom was being irrational when she got hers checked out by a doctor!
The odds of my house burning down are at least as miniscule as the odds of any random young black male being a criminal, so I guess I’m stupid for having fire insurance. I’ll cancel it right away, Matt!
So on and so forth. One might also invoke the common airline policy of not allowing adult males to sit next to unaccompanied minors. Not all adult males are pedophiles, but most pedophiles are adult males. Betting that a given adult male is not a pedophile isn’t worth the risk, and is certainly worth the minor inconvenience of asking all adult males to switch seats in the event they are seated beside a young girl traveling alone. Liability issues, you know.
And that’s what this second point is really about: the logic of damages. In a purely decontextualized way, I know that Pr(A|B) is not the same as Pr(B|A). Given certain context and content, however, betting on Pr(B|not A) may carry serious consequences if, however unlikely, I turned out to be wrong. In other words, taking minor measures to avoid the risk of certain events or probabilities, however unlikely, is not at all illogical. In fact, in many contexts, we commend people who take these measures. We call it “planning ahead” or “preparing for all possible outcomes” or “erring on the side of caution.”
This all seems to me a perfectly rational rejoinder to Yglesias’s over-simplified evocation of Bayes and its implications for how Zimmerman acted. Zimmerman’s neighborhood had recently been burgled by young black males, a small percentage of the community’s population. Trayvon Martin was a young black male, which, of course, did not mean he was a burglar. However, the possibility that he was a burglar, however unlikely, was nevertheless, given demographic reality, modest enough to warrant a few minutes of Zimmerman’s time to check him out, because not checking him out (i.e., assuming that Martin was not a burglar) might incur greater damages to the neighborhood.
Are you not rationally persuaded?!??!!?!
Nope. Not Aaron Gross, one of Sailer’s progressive trolls:
the fact that young black men are disproportionately likely to be involved in violent crime in no way licenses the inference that you should stop random black men on the street….
That’s absolutely correct. There may be other facts that imply it, but the facts that Cohen cites do not support it in any way. And it’s obvious that Cohen did get his conditional probabilities backwards, just as Yglesias pointed out. Cohen was talking about P(black|criminal), where a more relevant probability is P(criminal|black).
I haven’t read all the comments here, but from the ones I’ve read it seems that as usual, iSteve readers endorse the stupid, unsound argument over the intelligent, sound argument because the former supposedly leads to the desired conclusion.
Now, there may be a purely rational case against the one I’ve presented against Yglesias, another round in a logical tête-à-tête. However, Aaron Gross’s ostrich response is the most common one I’ve seen from the Left.
What am I to do with that? The “stupid” comments to which Gross refers lay out the same argument against Yglesias that I’ve provided here. To me, it seems like a perfectly reasonable counter-statement to Yglesias. To Gross, however, it’s the usual, unsound stupidity of biased proles grasping for rationalizations.
We’ve now reached the reason I brought this whole thing up: the incommensurable gap between me and Aaron Gross demonstrates that logic and reasoning are, in the end, absolutely worthless when it comes to arguing about social and political issues. It seems we cannot even decide on what constitutes “sound” versus “unsound” argument—on what constitutes rationality versus stupidity. More and more often, I find myself persuaded by Heartiste’s preferred tactics for cultural engagement: mockery and agitprop.
Which brings us back to the second sentence I wrote in this post: I agree with James Goulding that our blogs should elevate the discussion. But how can we elevate the discussion if the other side will never see as rational what we see as rational? What’s the point of trying to convert the Brahmins if they deride even a rational counter-statement as “stupid” and “unsound”?
At this point, I’m failing to see a point.
RELATED: Perhaps I’m being too pessimistic. William Saletan changed his mind about the Zimmerman case. (But then, Saletan is probably just a closeted reactionary.)