Specific-general inference, & bias

Some of the reaction to the shooting of Gabrielle Giffords looks to me like Jack Straw’s comments about some Pakistani men targeting white girls. Both are examples of drawing general inferences from specific events. We might call this the columnists’ heuristic, in deference to their habit of using single anecdotes as somehow representative of a wider truth.
So, for example, Paul Krugman blames the shooting on the “climate of hate” fostered by the far right. Michael Tomasky links it to “violent rhetoric”, whilst Dave Osler ponders the role of violence generally. Likewise, Lord Straw sees a few sex offenders and links them to the “Pakistani community.”
In both cases, however, there is an alternative possibility - that what we have are just random events and rare wrong uns.
What worries me is that  a combination of cognitive biases can lead us to fail to see this, and instead to infer too much from specific cases. I mean:
1. The confirmation bias. Paul Krugman always hated rightist rhetoric. Jack Straw always had his doubts about Muslim culture. Both are quick to find confirmation of their priors.
2. The availability heuristic. It’s easy to over-rate the prevalence of events if one or two of them get publicity, and thus to exaggerate the extent of a problem. This is not happening in the Giffords case - everyone knows it’s rare for Congressmen to be shot - but it might be happening in the case of Pakistani sex crimes, where statistics are less well known.
3. Outgroup biases. If a member of another tribe does something bad, we tend to regard such behaviour as representative of that tribe, whereas if one of our own tribe does it, we regard it as a mere idiosyncrasy.
4. Seeing patterns where this is in fact randomness. The technical term for this is pareidolia. But we could call it the “Deal or No Deal effect”; players on that fatuous show often seem to think there are patterns in what is in fact a random distribution of prizes. This leads us to look for causes even where there might be none.
5. A representativeness effect. We tend to think a causal relationship is more plausible if cause and effect resemble each other. So we naturally think that violent rhetoric provokes violent acts, or that a primitive attitude towards women leads to more sex crime. But we need more than mere likeness to establish causality.
Here, though, is the problem - I use the word “might” for a reason. In some cases, the columnists’ fallacy can be right. Individual instances can sometimes draw our attention to widespread problems - as Warren Buffett hinted it in his famous aphorism, “there’s never only one cockroach in the kitchen.”
The question is: how can we tell? One way is to ensure that we have plausible mechanisms. For example, in Buffett’s case, single instances of a company doing something badly can be a sign that management’s control systems are imperfect. Another way is to be more fastidious about the evidence. If it can be shown that disproportionate numbers of right-wingers attack left-wingers, or that disproportionate numbers of Pakistanis are sex offenders, then Straw and Krugman might have a point. In the absence of such evidence, though, what we have is mere hypothesising. But anyone can come up with hypotheses.

Comments

[nm]

Surely the likes of Krugman are basing their conclusions about this shocking incident on an observable deterioration in the debate in the US rather than just one incident. The "we are mad and we are armed" tendency in the US has seemed quite threatening for some time with the "Kill him" shouts about Obama at Palin rallies being a rather chilling example.

[chris]

Yes - it might well be that the nature of political discourse in the US has gotten more violent, at least relative to the 80s and 90s (though not the 60s - civil rights and Vietnam provoked strong comment).
But whether this is a cause of political violence is moot. I mean, we have (for now) only one instance of this.

[FDUK]

"If it can be shown that disproportionate numbers of right-wingers attack left-wingers, or that disproportionate numbers of Pakistanis are sex offenders, then Straw and Krugman might have a point"

Try looking here to see that this is not a one off incidnet

http://www.csgv.org/issues-and-campaigns/guns-democracy-and-freedom/insurrection-timeline

[BenSix]

I'd bet criminal psychologists get really narked at times like these. I mean there they are, desperately trying to map the psyches of these murderers and yet, with little more than someone's reading list and a bone to pick, commenters from Krugman to some bloke on a U.K. weblog know the score before the gun has even cooled.

[Ourben]

"If it can be shown that disproportionate numbers of right-wingers attack left-wingers, or that disproportionate numbers of Pakistanis are sex offenders, then Straw and Krugman might have a point"

To compare what Krugman and Straw said is a shocking false equivalence. Jack Straw drilled down to a quite specific demographic. Straw referred to specific mores of a specific sex in a specific town.

Pakistanis are not all men. They're not all in Blackburn. They're not all unmarried and expected to marry a girl from Pakistan.

I'm having a hard time reconciling the differences in editorial standard between this article and the last...

"Straw sees a few sex offenders and links them to the “Pakistani community.”"

No he did not. The truth is, we do not know what he's seeing. It is you that is making the link to Pakistani community. I understood immediately that he was referring to a very specific group perpetrators and a very specific group of victims.

Your mind. Not his words.

All we know is that the data on public record doesn't support his hypothesis. There may be undisclosed data that explains his concerns, there may not be.

People used to talk about a specific problem within the Catholic church. The abuse of especially vulnerable young boys. The Vatican refused to release documents from internal investigations. For years people speculated with no evidence. After time many victims have come forward.

Would things have been different if we had given more credence to the allegations, asked the Catholic church to be open and we'd all made an effort sooner to talk about it?

It's not a difficult question.

[Laban]

To be fair to Jack Straw, the existence of grooming seemed to be a shock to him in 2006.

http://www.lancashiretelegraph.co.uk/news/819890.girls_groomed_for_sex/

Blackburn MP Jack Straw today expressed his deep concern and said he would be raising the matter immediately with the Home Office.

He said: "I was frankly unaware of the scale of this problem until the Lancashire Evening Telegraph highlighted it but it is one that we all must now address including the Asian community."