Cliodynamics, The Hard Sell

"we must transform history into an analytical, predictive science"

That's absurd. Human dynamics is chaotic, and more chaotic than in the past.

I'm reminded of "If you're an economist, how come you ain't rich?"

"Psychohistory" was a field practiced in Isaac Asimov's science fiction books but as I recall in the novels its practitioners had the advantage that they could actually read other people's minds, also.

One mistake the Foundation made was to assume that following unpredicted events, like The Mule, the system would return to some equilibrium track per the equations. Chaotic systems don't work that way, and they are their own unpredicted events.

Looks to me that the great shifts in society are almost never predicted. The stock market has anti-prediction forces; maybe social systems do too.

There's a Korean action/drama where the main guy is an engineer (unusual for an action-thriller kind of show) who becomes incidentally involved with a woman from the future the Big Bads in the present are after for <insert reason>, one among many they call "illegal immigrants."

Later he learns that he's actually the guy who invents the time machine and is the only one who knows, or will know, how to do it.

And the Big Bads catch him repeatedly and he repeatedly escapes, not just because he's that I-can-turn-15-different-household-objects-into-a-bomb kind of character, but because they're finally too afraid to kill him, as they really don't know what happens if they do that. and at the very least make their "immigration control bureau" position a superfluous government job.

Is that a TV show? I'm glad I don't watch TV.

It is! I read somewhere that all fiction, time travel or no, at an abstract level are the same six basic stories.

In more pedestrian news related to history as a "predictive science" Coca Cola shares have risen 2% since the GOP announced a boycott of the company (God, what did Coca Cola do now, attack the Capitol building or something?), showing how seriously the market seems to take the GOP's ability at convincing Americans in large numbers not to buy Coca Cola products.

Right. Directly analogous to this:

"You can't predict the market, because the market is a prediction".

Any new analytic technique for planning the future will actually change the future, but in an unpredictable way. However, there will still be a power imbalance between those using the new analytics compare with those not.

CH

We take a flask of rum when we go skiing, to make rum+cokes with lunch. Northstar only had Pepsi. It worked fine.

Good idea! We already have a name for it - it's called "Hindsightology". :)

We used to call it curve fitting. Now we use a zillion lines of buggy code on a supercomputer instead of a simple polynomial.

Curve fitting isn't limited to curves defined by polynomials. When I did it for my Ph.D., work back in the late 1960's, the curve I was fitting was defined by the numerical integral of a differential equation - I also ran a variant where the curve was defined by the numerical integrals of three simultaneous differential equation (for concentration, pressure and temperature) which turned out to be a more computationally expensive way of getting the same result, and not worth the extra effort. The IBM 7040/44 on which I ran the code certainly wasn't any kind of super-computer. Seymour Cray was working for Control Data Corporation at the time, and didn't start building the Cray super-computers until 1972.

I used the Fletcher-Powell approach. It's still around as the Davidon-Fletcher-Powell Method. Davidon seems to have published the basic idea in 1959, and Fletcher and Powell made it work around 1963.

If you want to be rude about the imagined defects of curve fitting, you need to learn a little bit more about the subject.

It's certainly easier to fit curves to orthogonal polynomials - Fourier transforms are a case in point - but real data doesn't always lend itself to such a decomposition.