does non-falsifiable imply not true?

This post is a longer-form discussion following this Twitter thread. The thread began with Steve Vaisey expressing interest in how gender scholars would respond to this article, which apparently shows that women in more-gender-egalitarian societies have personality profiles more different from men than do women in less-gender-egalitarian societies. It then presents evolutionary psychology as one way that people might interpret that finding, implying that gender-based personality differences might be “natural,” not socially constructed, since they are more different when society “gets in the way” less, i.e., when society is freer.

I have still not read the article, but only the abstract, so my comments are about the discussion that followed, not about the quality or interpretation within the study.

First, let me recap the Twitter thread about more-free societies. The study’s assumption is that more-free means less-society, i.e., traits observed more in more-free societies are more likely to be “natural,” in the sense of pre-social, traits.

Any reasonable theory of social construction must include the fact that preferences and desires are the result of iterative processes of social construction. You can get this from any garden-variety psychoanalytically-informed theory, from Freud to Chodorow to Foucault. Add to this the fact that active repression can produce seasoned resistance (e.g., here ), and it is no longer reasonable to talk about societies as being unidimensionally more- or less-free. Thus, there are plausible genealogical accounts in which “more-free” societies would predict less variation and plausible genealogical accounts in which “more-free” societies would predict more variation.

I don’t see a way around this as a faithful statement of the theory. But it produces an epistemological mess. Steve is “astonished“; Nicholas Christakis worries:

I sympathize, and I do hope to grasp the truth about the world. I like falsifiability, and I think it’s a crucial tool for disciplined inquiry. But at least at this scope, the theory as I’ve outlined it is unfalsifiable, since the same cause predicts opposite effects. I actually think one could imagine more traction by moving to a lower analytic level, for example by trying to characterize the types and histories of (non-)freedom rather than trying to characterize entire societies as more or less free.

But assuming we can’t do that, we’re left with this conundrum. Posit, for the sake of argument, that the claim I’ve outlined above is ontologically true. That’s at least plausible. And posit that it’s also impossible to discern the truth of that claim because of the epistemological problems already discussed. Logically, both of these can be true. That’s frustrating; it dashes our hopes to grasp the truth about the world (for this particular question); but it isn’t evidence that the claim isn’t true.

I think Steve is asking to update the likelihood of the claim being true based on the new information in the article (right, Steve?). Essentially, that means bracketing the complexity of the claim and figuring that there’s some validity to the study’s interpretation, even if that interpretation isn’t dispositive. Well okay, I’m willing to go with that, but only if we recognize that then the evidence in the article is similar to qualitative, historical, genealogical, and similar evidence: clues as to the operation of a causal process that is inherently unknowable.

But in the bigger picture, the fact that we don’t have the tools to assess a claim does not make that claim less likely to be true.

 

Author: andrewperrin

University of North Carolina, Chapel Hill

4 thoughts on “does non-falsifiable imply not true?”

  1. Interesting. FYI, I was disgusted to learn that Sci-Hub has stolen and distributed a copy of the paper you’re talking about, here: http://sci-hub.tw/10.1002/ijop.12265.

    I don’t care much for all that evolution stuff. But with regard to gender (which might not be what you really wanted to talk about), this is also the old debate about “horizontal” versus “vertical” segregation. The paper appears unconcerned with how “egalitarian” is measured, but some places with small gender gaps in pay have lots of segregation (Sweden, if I recall). Liana Landivar and I wrote the definitive review of Charles and Grusky, focusing on this point: https://osf.io/preprints/socarxiv/syde8. (Disclaimer, I’ve also written papers with unitary measures of gender inequality, so sue me.)

    Here are some relevant excerpts:

    Charles and Grusky argue that gender segregation is not necessarily eroded by the march of modernity. In fact, post-industrial development (which pulls women into nonmanual jobs) and liberal egalitarianism (which protects access to jobs based on individuals’ existing gendered preferences) together threaten to make labor market “hypersegregation” permanent. …

    Horizontal segregation is said to result from ideas of gender essentialism, while vertical segregation follows from cultural notions of male primacy. …

    Charles and Grusky make a strong case that, although liberalism erodes pay and hiring discrimination, essentialism remains a basis for occupational segregation: “The insidiousness of essentialism is that it clothes segregation in voluntarist terms” (p. 336). But, if essentialism only contributes to “horizontal” segregation, why is it insidious? To avoid that pitfall we need a more gendered notion of status. The terms “vertical” and “horizontal” conjure axes on a plane defined by an economic outcome (socioeconomic status). But what about gender itself? We need to uncover gender’s own hierarchical logic, not just its effect on economic rewards. And in the autonomous logic of gender, essentialism and primacy are not so easily separable.

    I guess by my logic, you could say that maybe societies with high gender “egalitarian” scores but large gender personality differences aren’t actually as egalitarian as your measure implies.

    Liked by 1 person

  2. I wonder to what extent there is a spurious relationship here impacted by third variables. For instance, perhaps more capitalist societies also tend to be more ‘free,’ and also capitalism reinforces gender differences because if there are more gender differences, then twice as much stuff can be sold to parents who have a boy and a girl.

    Like

  3. Andrew P.:

    Regarding your statement, “But in the bigger picture, the fact that we don’t have the tools to assess a claim does not make that claim less likely to be true,” I think you’re making a big assumption here, which is that the sorts of claims that are made in social-science articles, and in publicity for and discussions of social-science articles, can be characterized as true or false.

    It’s my impression that most of the claims we see flying around are too slippery to be true or false: at one level, they’re vague enough to be true automatically (“evolution affects behavior,” “people respond to incentives,” etc.), at another level they correspond to statistical statements regarding particular datasets without any clarity about what is the general claim that could be evaluated.

    That’s ok–not every claim needs to be clear. My point is that it can be a mistake to try to identify claims as true or false.

    Like

Leave a Reply to Andrew Gelman Cancel reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s