iWAR: The Weird Analytics Review

Posts Tagged ‘false positives

Political Correctness, Ideology, and Bad Analysis

leave a comment »

Politico came out with a piece a couple weeks ago titled “Right fractures over Islam.” The author of the article quotes David Horowitz at CPAC:

“We are also faced at home and abroad with a mortal threat in political Islam. Political Islam is a totalitarian movement that seeks to impose Islamic law on the entire world through the seizure of states by stealth and electoral means where possible and by terror where necessary and sometimes by a combination of the two. There are hundreds of millions of believers in political Islam.”

Apparently CPAC featured a panel on “the threat of sharia law” as well as on the “ground zero mosque.” A small panel discussion about religious freedom moderated by Suhail Khan, the only Muslim board member of the American Conservative Union, was interrupted repeatedly with questions about Khan’s possible ties to or sympathies toward (and his parents’ ties to or sympathies toward) the Muslim Brotherhood.

Khan pointed to activists Pam Geller and Jihad Watch’s Robert Spencer as two prominent proponents of the idea that radical Islam is the enemy and that many, if not most, Muslims are suspect by association. The two paid for one panel at CPAC where “concerns about Islam itself as a faith were openly voiced by both audience members and panelists.” For example:

“For 10 years, people have been asking for moderate Muslims to speak up,” said Spencer. “We’re going to be waiting for those guys until doomsday.”

“Moderate Muslims don’t exist,” said one audience member at the Geller and Spencer event. “Muslims are not able to be moderate — or they are speaking against what is written in the Koran.”

Now, for the sake of argument, let’s assume that all that rhetoric isn’t just political posturing. I think that’s a fairly safe assumption since there appears to be a lot of people who look at connections between Islam and extremism without, apparently, ever trying to score political points from the endeavor (see here for an example). For the sake of convenience, I’ll refer to these kinds of connections as ideological explanations of behavior.

I’m opposed to ideological explanations of behavior, not because they aren’t politically correct, but because they represent really shamefully bad analysis.

It’s impossible to connect ideology to extremism on the basis of empirical observation alone. To do that, we’d have to see that a lot of people who espouse a certain ideology engage in the behavior that concerns us (in most cases, violence). Rhetoric like the stuff seen at CPAC show that the case can be made that we do, in fact, see that connection between ideology and behavior. However, those kinds of observations aren’t enough to claim that we have evidence of a connection. We would also have to show that many people who do not espouse the ideology do not engage in the behavior, and that few people who espouse the ideology do not refrain from the behavior. In other words, our assertions about cause (ideology) and effect (extremist behavior) need to create few false positives and false negatives.

Ideological explanations of behavior don’t do that. Even if we define “Islamist” ideology ridiculously narrowly (violent salafist jihadist” or whatever), those definitions invariably result in false positives and false negatives – people who do not engage in the behavior despite their espousal of an ideology, and people who do engage in the behavior despite their ignorance or even rejection of an ideology.

Now, we don’t technically need a false-positive-less and false-negative-less  set of observations to connect a purported cause and a purported effect. Smoking causes cancer but if we use the fact that people smoke to predict who is going to die of cancer, we’re going to end up with a lot of false positives and false negatives, because a lot of other things can mitigate smoking’s carcinogenic effects or have carcinogenic effects regardless of whether a person smokes. However, in cases where effects are determined by so many interacting causes that a clear empirical line cannot be drawn between an individual cause and an individual effect, good analytic practice demands that we define actual mechanisms by which the purported cause can produce the effect.

Mechanisms aren’t just stories. It’s easy to say something like, “If a person has an ideology the defines one group of people as an enemy, then you will be more likely to be hostile towards members of that group and less likely to feel empathy with those people.” Sound very plausible. But what reason do we have to believe it? I have yet to see an attempt to define a mechanisms by which ideology elicits behavior that goes beyond these kinds of just-so stories.

Fact is, we have no reason to believe that ideology influences anything at all. We can talk about ideological influences that seems to make intuitive sense, but intuition is a lousy way to conduct analysis. The Islam-related rhetoric that came out of CPAC ranged from politically incorrect to plain-old bigoted, but that’s largely beside the point. The point is that it’s bad analysis.

Written by Schaun Wheeler

February 28, 2011 at 8:48 am