Thread by @MetacogniShane

I keep seeing this study shared lately as some sort of proof that "people who don't socially distance are dumb! lulz." It's been hyped as supporting important link btwn working memory capacity and social distancing.

Let's look at why it really doesn't... pnas.org/content/early/…

Let's look at a specific claim by the authors, that their results show that WM capacity plays a key role in social distancing and possible route for intervention strategies...

The good: The study is well-powered to find an effect given their analytic approach (N = 1031 after exclusions). They also used a diverse battery of personality, dispositional, and intelligence measures. So far, so good....

They posit that many people may not be socially distancing b/c they have lower capacities for holding & manipulating multiple pieces of complex info (i.e., WM) in their heads, inhibiting their ability to sufficiently weigh costs/benefits of engaging in social distancing....

Using a multiple linear regression model, they find WM is a significant predictor of self-reported social distancing compliance. This is all accurately reflected in results. However, a closer look shows that: 1) 4 other predictors had larger effect sizes than WM, and...

...2) their final model accounts for only 22% of the variance even though it has 7 sig. predictors. In fact, adding WM to model only increases R^2 by .02, suggesting it's only contributing 2% of explained variance in soc. distancing (depending on how you interpret R^2).

In fact, out of the 7 predictors, 4 had larger betas (effect sizes) than WM. These include, age, gender, depressed mood, anxious feelings, and agreeableness. In fact, openness beta (.13) very close to WM (.14).

So, why the extreme, obvious focus on WM as a predictor, as if it is somehow theoretically more important than the other variables with larger effects? And why gloss over the fact that the overall model still fails to explain the other 78% of variance in scores?

One key factor I feel they left out is political affiliation and I might also throw in news consumption behaviours. Until recently, major Republican political leaders and pundits discouraged masks and social distancing from the start.

axios.com/political-divi…

My hopefully educated but possibly dumb guess would be that political affiliation (among other things) should have been included as it would likely explain as much, if not more, variance than WM. Huge focus on WM results here seems misplaced, to put it kindly. End thread.

To be clear, I don't doubt WM plays some role (also analytic reasoning processes), as it does in virtually every information-processing-based decision but its influence appears small, so WM interventions likely ineffective.

I accept that I might be totally wrong tho.