11 September 2017

Are we looking at the end of science?

So, "gaydar" is a real thing.

Rather than read all the nonsense, go straight to the authors' statement first, then look to those who would debunk. Now, lest you think me a homophobe for using the word gaydar, I've been busted -- mocked might be the better word -- by my gay friends for my gaydar not being very good. They also mock me by saying I should look into Garanimals so I can go into public wearing clothes that match and there is this 18th century technology called an "iron" that could help with the rumpled look. There has been some angst in some of the LGBTQ.* community about the study, but they're wrong for all the wrong reasons (cf. authors' statement), what's really at stake here is the implication that there is a correlation between "behavior" -- in this case sexual preference -- and physical traits.

Before you get all bowed up about where I'm going, go away and listen to the Radiolab podcast on Alex the Gondolier. People are people and have the right to the pursuit of happiness. The best way, if imperfect, is to lead with equal protection for all under just laws that allow folks to live how they please as long as they don't unduly impinge on the rights of others. Equal protection for all. All. Got that? All.


The notion that we simply should not study certain things is not out of the mainstream. One of my very favorite public intellectuals, John McWhorter whom I've read with great interest and pleasure for years made the case to "stop obsessing" about evidence for race and IQ correlations, that it serves no purpose, in National Review. Read it by all means, and be sure to follow the link to the excellent reference he makes to "On the Reality of Race and the Abhorrence of Racism".

The problem is, that in modern society where outcomes are increasingly correlated to IQ, g, cognitive horsepower, whatever, in order to enable all our citizens the best shot at life, liberty, and the pursuit of happiness -- a goal I hope you won't find too cretinous -- perhaps we need to structure our values and programs toward actually achieving that goal as opposed to conflating the means with their ends. This is pretty much why Charles Murray got lynched for saying in The Bell Curve. That we should care about people. I won't link to SPLC or other slander, but you can google if you like about the lynching bit. Read what he actually writes, listen to what he actually says, then decide for yourself if he's all of those horrible things that have been said about him.

So, if g is heritable (relatively uncontroversial), and other stuff is heritable like height, eye color, whatever, what happens when we turn machine learning algorithms on correlating g and physical traits that can be gleaned en masse from Instagram, Facebook, oh, the DMV, etc. What if the correlation is between other behaviors we don't think as necessarily heritable or caused by prenatal hormone levels, etc? What happens if algorithms can pick out, I don't know, you name it, probability of rape accusation with 70% accuracy? 55% accuracy? I am not endorsing physiognomy writ large -- I have never really much cared or thought about it -- but what if the evidence turns out to support it? The gaydar thing works pretty well, it seems.

It cannot be argued that much of this information would be useful in a wide variety of contexts. But it cannot be argued that the potential for misuse is tremendous. Here is the problem with population statistics: they're incredibly useful about saying things about populations, but they really can't tell you anything about any individual.

To illustrate, let's play a game. You get 10 pennies and I get 5 six sided dice. We're going to play a game where we put in matching antes for each roll, let's say $100. Whoever gets the larger sum (tails 1, heads 2, and each pip on the die counts as 1), takes the pot. My minimum low at 5 is lower than yours at 10, but my maximum 30 is higher than your 20. We're going to play, oh, 1000 times. 10000 times. A million. I don't care. Who wins roll 42? Who knows? But you'd never play that game. Um, unless you do, let's get in touch, because I'll play all day long. Night, too, even if it means coffee after 1500 or so.

If you can play enough games, it's enough to flip 10 pennies against 9. The thing is, if the difference in the odds is detectable, then it's exploitable, but it's also, perhaps, understandable.

We need to press on with the science and let the evidence lead us where it may. But, so too, do we need our shared mythology, or belief, in the the right of all people to live their lives. Science is, in my opinion, our best and most effective collective reasoning construct we have created (beats reading entrails or thrown knucklebones or thinking that Congress can understand the definition of hypocrisy) and one we cannot afford to lose, even at the cost uncomfortable truths. But this is a really big deal, and I'm not sure this is a conversation we're collectively ready to have.

No comments:

Post a Comment