Monday, January 10, 2011

Impossible demands for "proof" in the Giffords assassination attempt

Louisiana State University


Although it was an utterly predictable response, I was still disappointed at the re-emergence of one aspect of the “lone nut” response to the Giffords assassination attempt: the erection of an impossible standard of proof in linking the alleged shooter to violent rightwing rhetoric and imagery. As if you can’t talk about the relation between violent rhetoric and violent acts without showing some direct causation. 
But what can you expect from blog comment threads, I told myself. Fighting the misguided and the trolls might give you an insight into the median winger blog commenter, or at least those brave enough to take the fight to the sort of leftie-liberal blogs I frequent, but it’s not going to do much more than give you a break between serious reading and writing your own work. 
But then I read this post over at Lawyers, Guns, and Money, and I realized that some social scientists, in trotting out that old show pony, the distinction between correlation and causality, might have a really impoverished view of “causality.” 
Here’s the post: 
January 10, 2011 | Charli Carpenter 
John Sides at the Monkey Cage weighs in with some social science on the relationship between militant metaphors in political speech and individuals’ willingness to engage in actual political violence against government officials. The findings he cites: an experimental study has shown there seems to be no effect on the overall population of exposure to “fighting words” in political ads, but there is an effect on people with aggressive tendencies. Moreover: 
This conditional relationship — between seeing violent ads and a predisposition to aggression — appears stronger among those under the age of 40 (vs. those older), men (vs. women), and Democrats (vs. Republicans).  


But his real point is that we should be cautious of inferring from this or any wider probabilistic data causation regarding a specific event: 
To prove that vitriol causes any particular act of violence, we cannot speak about “atmosphere.” We need to be able to demonstrate that vitriolic messages were actually heard and believed by the perpetrators of violence. That is a far harder thing to do. But absent such evidence, we are merely waving our hands at causation and preferring instead to treat the mere existence of vitriol and the mere existence of violence as implying some relationship between the two. 
I left the following response at LGM: 
So that’s it, a binary between “hand waving” and billiard ball causality? Somebody’s got a terribly impoverished view of “causality” here. I’d say it was an example of “physics envy” but contemporary physicists aren’t that crude. 
Let me give an analogy to a well-known biological principle, Schmalhausen’s Law, to show that we can make sense of the interchange of environment and population w/o meeting an impossible billiard ball causality standard. Schmalhausen showed that in species-typical environments, developmental robustness hides a lot of genetic variation. In other words, in normal environments you can get roughly the same results in a population with genetic variance. But put that population under environmental stress and the previously hidden genetic variation shows up in a greater range of phenotypes. This is not “hand-waving” but neither does it adhere to an impossible physics-envy billiard ball causality standard. 
The analogy here of course, is that today’s political rhetoric environment is so extreme that we can plausibly suppose that it will expose the psychological variation in the population that would otherwise remain unexpressed. 
This is not hand-waving, and it shouldn’t be dismissed because it doesn’t match some ridiculous standard of a direct cause-and-effect of one statement to one act.
I went over to New APPS and made a post of it, where I got this interesting response by Scalinger: 
In fairness to Sides, and in keeping with your example: an extreme environment can “expose” variation only if members of the population are actually affected by it (by its extremes, moreover). So it's not out of place to ask whether the assassin ever actually saw or heard violent messages. I don't see any physics envy in that.
My reply: 
Yes, that's fair to ask. The exposure to the rhetoric part is easy, given what we know even now. The shooter was described as "obsessed" with Giffords, he attended one of her rallies in 2007, she won her election by 3500 votes against a candidate whose campaign had all sorts of violent images. It's a vanishingly small probability that he wasn't exposed at some point to these sorts of things. 
Now as to Sides' second requirement, "belief," we're a lot closer to unilinear causality than I'm comfortable with. We have to prove that he had a representation with the content "Giffords must be eliminated" and that we can trace that representation to an event or series of events at time T1…Tn (the exposure to a particular message or set of messages)? So that this representation with that content (plus some other representation) are then the necessary and sufficient conditions for his action? That's not physics envy? 
OK, maybe he has a more sophisticated psychology than that, though it's hard to tell from his post. And it's certainly no good on my part to just chant "nonlinear dynamics" as a mantra so that anything goes in linking environment and shooter. 
But there has to be something along the lines of dispositions and thresholds that's better for thinking this case than the sort of linear belief-desire-action scheme he seems to be proposing (what Susan Hurley memorably mocked as part of the "classical sandwich" view: sensory input -- computation on representatons -- motor output). 
In this nonetheless heartfelt piece, Jon Stewart displays the same exclusive binary to which I object: either we can show a straight line causality or we can't make any sense out of a "complex ecosystem." So it's either a coherent ideological act (message – belief – action) or it's "senseless," it's "lunacy." But there are other forms of causation than straight line efficient causality, for example, the sort of environmental pressure causing increased phenotypic variance that Schmalhausen's Law describes. I hope we can say something analogous about the relation between the Giffords shooting and the extreme political rhetoric and images of Tucson Arizona without it being mere correlational "hand-waving." And actually I think Stewart's "toxic environment" trope is a good place to start.
The Daily Show With Jon StewartMon - Thurs 11p / 10c
Arizona Shootings Reaction
www.thedailyshow.com
Daily Show Full EpisodesPolitical Humor & Satire Blog</a>The Daily Show on Facebook

To wrap this up, it seems to me that this billiard-ball-causality view has a lot in common with the isolated libertarian subject. Libertarians, I think, have to deny corporate advertising effects on the formation of choices, since that’s their bedrock, the individual and a consistent preference set. So they will deny any cultural influences and the porous subject that goes along with that. So there’s the isolated hard-shell individual (the billiard ball), and the only thing that will influence it, short of literally banging into it with a real billiard ball, is direct gun-to-the-head government coercion. Hence their refrain in pushing the “taxation is theft” line: "men with guns will come to your house and make you pay taxes!" This impoverished view of causality seems to be what’s behind many demands for “proof” in the Giffords case.

No comments:

Post a Comment

LinkWithin

Related Posts Plugin for WordPress, Blogger...