Seth

Admixture of neuroscientist, data scientist, and wife guy. Also writing about the implicit computational content of literature.

[Seth @fewmetz]

@sociologyWV You could, unironically, do it like The Big Short. You know, cut to Ishmael in the bathtub for 5 minutes while he explains the superiority of the sperm whale over the right whale. Maybe even cast him as Margot Robbie.

[Seth @fewmetz]

To the extent economists believe that events generally have a single cause, it is because their JMP is about a single variable and they desperately need it to be as important as possible.

[Seth @fewmetz]

This... doesn't make any sense. Econometrics exists because multiple casual variables impinge on any particular event. Events are overdetermined when most of those variables point in the same direction, which will be the case in a reliable and predictable system.

[Jesús Fernández-Villaverde @JesusFerna7026]

If you have ever ventured outside the comfort of economics and attended a seminar in the history or sociology department (or, if brave, the English or Comp Lit departments), chances are you heard someone call an outcome “overdetermined.” I have heard it more times than I can count, most recently this morning, reading (an otherwise excellent) history book.

The first time I ran into the word, I had no idea what it meant. I had never heard it in college or in graduate school. So, I googled it (this was in the dark ages before Claude) and learned that an outcome is overdetermined when it has more causes than it needs to produce the stated effect. The classic example is two gunmen who shoot the same person in the heart at the same instant: either bullet kills them.

Ah, I shouted, I heard that example in criminal law I, when I was in law school (yes, those two incredibly boring months discussing causalism vs, finalism, for those who endured the course being taught from the German tradition)!

I also learned that the word has a distinguished pedigree. Apparently, it was Freud who coined it, Überdeterminierung, to describe how a single dream or symptom can be driven by several unconscious causes at once.

Althusser later carried it into Marxism, arguing that social contradictions are “overdetermined” because they pile up from many other contradictions rather than reduce to a single economic base. From Althusser it spread through critical theory, which embraced it because it is so handy for the project: “overdetermination” lets you deny that any social phenomenon has a single cause, and above all that the single cause is economic. You sound terribly sophisticated without the need to pay any real cost for it.

Here I had a good hearty laugh. The people who reach for “overdetermined” are often the same ones who tell you that economics is naively deterministic, that economists “believe in causality.” Yes, I have been told by my colleagues in other departments that believing in causality, or its twin brother the counterfactual, is one of the deepest sins of economics (I always wonder: if you don’t believe in counterfactuals, why did you get the COVID vaccine?).

But economics is the discipline that treats causality as an expensive medicine, to be dispensed with care. We pour enormous effort into identifying one clean source of exogenous variation, one counterfactual we can defend. We lose sleep over problems with far less glamorous names: selection bias, simultaneity, multiple equilibria, confounding, robustness. And very often we surrender and admit that we cannot identify the effect. Many times, I simply say: “I don’t know.”

We avoid “overdetermined” because if the outcome would have happened anyway through another channel, there is almost nothing interesting left to say about it, unless what interests you is rhetoric rather than substance.

In other words, the academics who accuse economists of naive causality are the ones deploying a word whose whole appeal is that everything causes everything. And they take that to be the more philosophically sophisticated position.

Well, at least “overdetermination” is not as silly as “historicizing the discipline.” Who came up with that gem?

[Seth @fewmetz]

"Very bad luck?" Why "very bad"? In a society with *any* kind of mobility, you will have downward mobility simply due to regression to the mean.

[T. Greer @Scholars_Stage]

I do think, however, that there is a very large class of people who grew up in the top 15%, and feel entitled to the sort of lives they grew up with, but either 1) lack the ability to maintain their class position 2) are simply disgusted with the kind of job/lifestyle that would allow them to maintain it 3) encountered very bad luck.

[Seth @fewmetz]

This is all fine, but shouldn't any discussion of "voluntary downward mobility" begin with regression to the mean? There are simply not a lot of life paths that are guaranteed to land you in the top income decile, and those paths will be congested.

[Kevin A. Bryan @Afinetheorem]

Love this but would go further: we have a class of "voluntarily downwardly mobile" esp in media/academia/culture/nonprofit. "Oyster farmer" sent to Hotchkiss by lawyer Dad and famous grandad guided by staffer who is Yale Law grad/grandson of Toys R Us founder: too on the nose 1/3

[Seth @fewmetz]

@Scholars_Stage I think there are a tremendous number in group 2 who are perfectly happy to acknowledge this tradeoff, but you do not hear from them because they are basically happy with the tradeoff they have made and therefore see no need to overthrow society.

[Seth @fewmetz]

@littmath I think it's probably that they don't really "want" anything beyond maximizing their relatively straightforward objective function.
Humans are just crazy non-stop thing-wanters, we'll want twenty contradictory things before breakfast.

[Seth @fewmetz]

@jessesmithsoc Ehhh this kind of thing will be painfully sensitive to the exact wording and prompting procedure. If the model can tell that you want it say the literature is bad, it will be happy to tell you that.

The value is probably more in parsing specialized jargon.

[Seth @fewmetz]

@joefrancis505 @RiccardoIorio4 I was responding to point b! Having consistently robust empirical findings would put economics well ahead of other social sciences!

Does the "general equilibrium model" part imply they are not described accurately?

[Seth @fewmetz]

@nabeelqu @ben_golub @littmath If you actually care about people taking AI capabilities seriously, then engaging in misleading hypesmanship seems like the worst thing you could do.
If people are unaware of AI capabilities, it is not for lack of hype; it is because the hyperatti are not credible.

[Seth @fewmetz]

@joefrancis505 @RiccardoIorio4 I think if the empirical results are actually robust and described accurately, with a high probability, that's a tremendous win for economics relative to other social sciences.

[Seth @fewmetz]

https://t.co/EIBUBFB2Pv

[Seth @fewmetz]

@krishnanrohit It's interesting that the text is PERFECT much of the time, and then sometimes just falls apart. Almost like there is a dedicated typographer in the model somewhere, but the final pixel-drawer doesn't always listen to it.

[Seth @fewmetz]

@JamesBland_Econ @ryancbriggs Yeah, if you've read anything about Stan at all you've read about informative priors and vectorization, etc. I suspect AI is more likely to struggle when things go "off-script", as they usually do at some point in any research project.

[Seth @fewmetz]

@Afinetheorem Fair enough, though I think people are far too eager to jump to most interesting question over the most practical one!

My own boring take is that the moral stakes are not actually all that high, and come mainly from the threat of people being weird about it.

[Seth @fewmetz]

@soumitrashukla9 @Afinetheorem Nobody knows. The upside is that, in 30 years, your kids may be represented in the dataset of an economics thesis trying to answer that question.

[Seth @fewmetz]

@Afinetheorem 2) The most rigorous work here compares siblings within the same families. The effects in that setting are consistently much smaller than in comparisons across the whole population. What this means "big picture" is not clear, but it is obviously relevant to embryo selection.

[Seth @fewmetz]

The galaxy-brain take is that it is pretty because it is brushstrokes!

[Andrew Fleischman @ASFleischman]

@TexasLawhorns You can think a painting is pretty even if you know it's all just brushstrokes