Tralfamadorians are four-dimensional alien beings able to travel anywhere in time as well as space. Or so Kurt Vonnegut reports, quoting one as saying:

"I am a Tralfamadorian, seeing all time as you might see a stretch of the Rocky Mountains. All time is all time. It does not change. It does not lend itself to warnings or explanations. It simply is. Take it moment by moment, and you will find that we are all bugs in amber."

My question for the day: could a Tralfamadorian make predictions? Short answer: yup, totally. Longer answer -->

Our root for the word 'prediction' wouldn't make sense to a native of Tralfamadore. It literally means pre-show - to "say or estimate that (a specified thing) will happen in the future".

This has always seemed a bit odd to me, only half of how we actually use the term. Looking at things from a Tralfamadorian perspective makes this more obvious: the word 'forecast' has no meaning on Tralfamadore. All past and future events are accessible to them. There's no such thing as a Tralfamadorian weather forecast. They don't bet on horse races. They can't try and game the stock market. They can't actually have a stock market.

But they could still make the kind of predictions we consider among the most important: what Gregor Betz calls an `ontological prediction'.*

Here are three famous ontological predictions (one mentioned by Betz). One: the existence of Neptune deduced from oddities in the orbit of Uranus (I see you snickering...). There was either another planet or Newton was wrong. It turned out there was another planet. Then two of Einstein's: light should appear to bend as it passes through gravity-warped space; and the existence of gravity waves. The first was famously (though not uncontroversially; see also this) confirmed by Eddington during an eclipse.** Confirmation of gravity waves is brand spanking new. Immediately, they are cosmically awesome, able to dig deep into the universe to solve the riddle of where some of the heaviest elements like gold come from (neutron stars colliding... whoooaaa).

So - those were all predictions, yes? And each did provide statements on the future - but only kind of by default. The future is the only place we can test our theories. On Tralfamadore, that's not true. Tralfamodorian Einstein could come up with his theories, look up a suitable eclipse in the seventeeth century on his four-dimensional road map, pop over to meet Newton for a bit of co-corroboration, maybe nipping to Papua new Guinea in 1698. (Probably best not to over-think it... wouldn't all Tralfamadorian predictions instantly propagate everywhere/when? So everything would by necessity be known instantly leaving nothing to be discov... oh, they're just a fictional device for making a point, OK then. Phew.)

All of which is a slightly belaboured way of saying: ontological predictions are fundamentally different to forecasts. They are timeless (though the realities they seek don't need to have always existed). They are about seeing things that were already there but we didn't know to look for.

The fact we have to test our ontological predictions in the future doesn't change how different this is to a forecast. It's unfortunate our definitions reinforce the idea that `predict' and `forecast' are the same. Of course, the two are dependent on each other: actual forecasts need underlying theory, and discovering that theory is an ontological-prediction job. But there is conceptual clear blue water between the two of them. New forecasts using an existing method don't require extra ontological prediction. You could also, for instance, improve weather forecasting independently of ontological prediction by throwing more powerful computation at it or some novel refactoring, without making any deeper discoveries about the underlying physics.

Why does this matter? Well, this is all a pre-amble to another post: after reading another attack on a cartoon version of Milton Friedman's argument about model assumptions, I feel like having a proper go at exploring why those (seemingly very popular) arguments miss the point, and why that's important. tl;dr: he never said "assumptions are irrelevant". He did say predictions are the ultimate arbiter - but it's hard to get very far without being clear what prediction actually is.

Once you start digging into this, it also ends up saying something about how different disciplines see themselves, how the public sees them, and how we frame the entire research enterprise in applied versus non-applied terms.

It's also a trope used by many people to reassure themselves that the entire edifice of economics is clearly stupid. That's annoying, wrong and that used to be me. But it's also used by people who should know better to bolster their economics-heretic credentials, and that's especially annoying. So, more at some point before 2019, I hope, or possibly before now if I can find a Tralfamadorian to work with. Thoughts gratefully received in the meantime: does this two-part distinction in prediction scan?

Update: oh, it turns out Friedman was quite explicit about including ontological prediction in his definition: "to avoid confusion, it should perhaps be noted explicitly that the "predictions" by which the validity of a hypothesis is tested need not be about phenomena that have not yet occurred, that is, need not be forecasts of future events; they may be about phenomena that have occurred but observations on which have not yet been made or are not known to the person making the prediction. For example, a hypothesis may imply that such and such must have happened in 1906, given some other known circumstances. If a search of the records reveals that such and such did happen, the prediction is confirmed; if it reveals that such and such did not happen, the prediction is contradicted." Source.

--
* Note Betz also thinks a prediction is a "statement on the future".
** Skipping over gravity having Newtonian effects on light - look, a black hole prediction!) Though if I'm reading that right, it's based on light being a particle with mass.