Saturday, April 30, 2016

The pseudoscience of liberal discourse gaming

The prospect of influencing and even manipulating public opinion has attracted some fairly rigorous and sophisticated intellectual inquiry for thousands upon thousands of years. You can find early traces in works as old as the Bhagavad-gita (in asides about how it is a "restless man's mind" that can be "strongly shaken") that extend in a fairly straight line towards modern scientific research (empirical experiments on how stress impacts amenability to persuasion, for example). En route, we've developed a fairly extensive body of knowledge about what works and what doesn't spread out over multiple fields: marketing, public relations, political science, psychology, and so on.

As in most fields of empirical knowledge, it turns out that many of our assumptions and historical ideas about public opinion are factually incorrect. To pick a trivial example, we now know that ancient rationalistic conceptions of humans as the "reasoning animal" are plainly false, and that people are afflicted with all kinds of powerful and irrational cognitive biases. These biases are often extraordinarily subtle, counterintuitive, and can't be recognized through sheer conjecture; usually, they can only be teased out through elaborate, carefully controlled experiments. If you study the literature, you'll understand how this works; if you don't, you won't understand it, and you're likely to conclude that people are rational in ways they are not.

So it's immensely frustrating to see armchair liberal discourse gaming, where we get sage advice grounded in theories of public opinion that just aren't true, and that no one in the field thinks is true. For instance, Jill Filopovic writes:
I think we've seen evidence that shutting Trump down fires up the GOP base & potentilly makes white voters sympathetic to him.
This is an empirical claim. What evidence? Think of how you would even substantiate this:
  • One thing you could do is just ask voters directly how "getting shut down" affects their attitudes towards Trump, and look for some unique reaction among white voters. But self-reporting on this kind of question is terribly unreliable methodology, and in any case no such polling has (as far as I know) actually been done.
  • More likely, Filopovic is relying on the related and enormously popular pundit methodology of relying on a personal sampling of anecdotal evidence culled from the self-reporting of various tweets and talking heads. This places all kinds of similar problems: while I'm sure TrumpTrainDad88 definitely said "This makes me want to support Trump even more!" there's no reason to conclude that his outrage is either reliable or representative.
  • Another thing you could do is look for any appreciable change in general favorability polling before and after such an incident. This has all kinds of serious problems too, however, since such changes are overdetermined and express all kinds of different factors - but hilariously, even if we set aside this problem, the polls actually falsify Filipovic's claim. For instance, prior to the Arizona protests, Trump's favorability was at 43% among white voters; that number dropped to 39% after the protest.
The fact is, if you look at the data on this kind of thing, what you find is people respond unreliably and inconsistently to disruptive protests. When Filipovic says "I think we've seen evidence" on this, I can't imagine that she's actually looked for evidence in any kind of rigorous or compelling way. (She's shown significant difficulty in understanding even basic polling in the past). What's more likely here is that she has an intuition about how the discourse works, largely informed by her conflict-aversion and her reflexive preference for polite rationalistic discourse, and that she will now backfill all kinds of anecdotal and statistically unsound "evidence" to substantiate this.

When liberals call for "an exchange of ideas on the left about the most effective ways to counter" opposition messaging, this is fine - but that exchange of ideas should be at least minimally informed by our understanding of public opinion and how it actually works. The notion that these ideas can be grounded in nothing more than ideology and personal preference makes such calls for dialogue little different than the Creationist calling for "an exchange of ideas" about the origin of species.