I loved this post and thanks for linking that Postman essay. I hadn't heard of it but I think that some of the same critique can be applied to his tradition that he applies to quantitative social science. Postman is a proponent of literary liberal culture and he's making the case that stories told in that idiom are more pleasurable.
But there's also a demand side to this story. It's not just that social scientists decided to start telling stories using numbers because they wanted to adopt the imprimatur of scientific rigor. We also live in a society where many non-scientific institutions reify quantitative metrics. The sports fans now demand those baroque advanced metrics. If “we’ve” “decided” that data journalism is more pleasurable than longform literary journalism, it’s much more likely that the stories that are told will be in that idiom
It's great. It also coins a really useful phrase regarding the overuse of mathematical models in policy contexts, for how such high-quant arguments provide "a patina of scientific and computational respectability".
"She could also not say—and had no aspiration to say—why it was that some children who
watched many violent programs did not act aggressively, or why some of
those who didn’t watch violent programs did act aggressively. Moreover,
she told me that within the past five years there have been more than 2,500
such studies conducted in American universities, with the result that there is
no agreement on very much except that watching violent television
programs may be a contributing factor in making some children act
aggressively, but that in any case it is not entirely clear what constitutes
aggressive behavior."
This passage is so good because nothing has changed since! There was the credibility revolution, but IVs don't work, and RCTs on questions like that have no external validity and usually produce, understandably, tiny effects because it's just 100% unclear how the causal mechanism works...
"She replied that [science] required one to be empirical, to measure things, to make one’s methods and conclusions public, and to test one’s assertions." This is exactly where we are today! Great post.
took me about six years of higher education in sociology and anthropology to reach Postman's conclusion; I wish I had been given that text on my first year! To be fair, anthropology has been much more self-critical than sociology, pol. sci., and so on; perhaps because of its colonial origins...
It's interesting that you lump in sports with "stories told with correlations". Are you asserting that there's no (objective) predictive power in ("moneyball" style) statistical modeling in sports?
I'm more referring to the random stats you get during First Take or SportsCenter: "Number of points scored in the third quarter of back to back games by players over 35 years old..."
I loved this post and thanks for linking that Postman essay. I hadn't heard of it but I think that some of the same critique can be applied to his tradition that he applies to quantitative social science. Postman is a proponent of literary liberal culture and he's making the case that stories told in that idiom are more pleasurable.
But there's also a demand side to this story. It's not just that social scientists decided to start telling stories using numbers because they wanted to adopt the imprimatur of scientific rigor. We also live in a society where many non-scientific institutions reify quantitative metrics. The sports fans now demand those baroque advanced metrics. If “we’ve” “decided” that data journalism is more pleasurable than longform literary journalism, it’s much more likely that the stories that are told will be in that idiom
I felt myself channeling Munger Theory as I wrote this post. And I think Postman would agree with you that much of this is demand side.
I'd never heard of this essay by Postman before, thanks for sharing. Along the same lines, I'll plug Kevin Baker's 2019 essay "Model Metropolis".
https://logicmag.io/play/model-metropolis/
It's great. It also coins a really useful phrase regarding the overuse of mathematical models in policy contexts, for how such high-quant arguments provide "a patina of scientific and computational respectability".
Agree, Kevin's essay is excellent. Highly recommended.
"She could also not say—and had no aspiration to say—why it was that some children who
watched many violent programs did not act aggressively, or why some of
those who didn’t watch violent programs did act aggressively. Moreover,
she told me that within the past five years there have been more than 2,500
such studies conducted in American universities, with the result that there is
no agreement on very much except that watching violent television
programs may be a contributing factor in making some children act
aggressively, but that in any case it is not entirely clear what constitutes
aggressive behavior."
This passage is so good because nothing has changed since! There was the credibility revolution, but IVs don't work, and RCTs on questions like that have no external validity and usually produce, understandably, tiny effects because it's just 100% unclear how the causal mechanism works...
"She replied that [science] required one to be empirical, to measure things, to make one’s methods and conclusions public, and to test one’s assertions." This is exactly where we are today! Great post.
took me about six years of higher education in sociology and anthropology to reach Postman's conclusion; I wish I had been given that text on my first year! To be fair, anthropology has been much more self-critical than sociology, pol. sci., and so on; perhaps because of its colonial origins...
It's interesting that you lump in sports with "stories told with correlations". Are you asserting that there's no (objective) predictive power in ("moneyball" style) statistical modeling in sports?
I'm more referring to the random stats you get during First Take or SportsCenter: "Number of points scored in the third quarter of back to back games by players over 35 years old..."