That word theory gets thrown around a lot. Some of my colleagues hold it to a really high bar whereas others use it pretty interchangably with hypothesis testing.
There’s an early phase of research that I’m not sure how to label. It’s not so much about levels, but something else. Here’s an example: what would you call the contribution of Copernicus to planetary motion? Ptolomy had these elaborate descriptions of everything revolving around the earth as cycles and epicycles to make up for wonky trajectories, and Copernicus came along and demonstrated that it all becomes a lot simpler if it’s all revolving around the sun. “Theories” of why the planets revolve as they do (Newton’s gravity and Einstein’s bending space time) came later.
Was Copernicus’s contribution a theory, replotting the data in a more sensible way, or something in between? Whatever it was, it was important, and it led to all that followed. But what do we call it (aka how do we regard it)?
@NicoleCRust To me, Ptolemy’s and Copernicus’s accounts of planetary motion are theories. A theory is a compression algorithm: a simpler representation of a mass of empirical data. An effective compression algorithm relies on discovering patterns amid the data, and so does a good theory.
If you say a shorter algorithm is a “better” theory than a longer one, you are saying something like Occam’s razor.
@NicoleCRust@JamesGleick I like your exchange of ideas very much! For me, a theory was always a (mental) model we put on the real world in order to make it both understandable („explanatory“) AND be able to create prediction/foresight. Is that a wrong/naive conception?
@JamesGleick@NicoleCRust I tend to view theories as approximations, each of which has predictive validity in some specific domain and fails outside that domain. So an early, relatively primitive theory is still a theory, just a first-approximation one. I'm still working on articulating to myself the distinction and relation between predictive power and explanatory, especially in the face of quantum theories with high predictive accuracy but disputed interpretation. Epistemology is knotty stuff
@mrcompletely@NicoleCRust Another way of looking at that is that all scientific theories (that we know of; in the world we happen to live in) are LOSSY compression algorithms. To achieve simplification, some stuff has to get left out.
Some physicists believe in the possibility of perfect theories, theories of everything, equivalent to lossless compression algorithms. I think that’s a wild-goose chase, in this universe, but I’m just guessing.
@JamesGleick@NicoleCRust I like that analogy a lot, because it's intuitive to me as a software professional that spends a lot of hobby time in areas that end up touching on compression algos. I tend to agree with your intuition about the hopelessness of the search for "losslessness" in theory but 🤷
The "increasingly accurate approximation/compression" view doesn't speak to domains, e.g. quantum vs relativistic physics, where each is highly accurate in its domain but breaks down in edge cases
@adardis@NicoleCRust I would begin by pointing out how much knowledge of our complex world is already embedded in the words “noises,” “ceiling,” “poops,” “hole,” “sack,” and “rice.”
What would the raw data look like if you didn’t have those concepts to organize them? Consider just the poops. You might start with some morphology; some (bio) chemistry. By the time you get to “poops,” you’re a long way toward the mouse.
@adardis@NicoleCRust To put it another way, your scenario is more a language puzzle than a scientific one. I suspect an LLM could come up with “mouse” more efficiently than a visiting alien physicist.
(I do like the conceit of the mouse as a theory for the data.)
@JamesGleick@NicoleCRust "There's a mouse" does provide a very compressed representation of a tremendous volume of predictions that an agent's theory makes about things: what it would look like if you saw it, how it behaves, etc. Without the concept of a mouse, mouse-generated data would be pretty puzzling to an alien physicist.
Would you be interested in writing about a 'new' compression algorithm for Aerodynamic effects, which regards the moving Aerofoil as an ENERGY DISTRIBUTOR...?
This 'theory' would logically form a shorter algorithm than 'proxy' analyses built around an assumption of 'Airflow', which is really the 'reverse model' found
@NicoleCRust I would actually call these all theories, in that they play (or played) the defining role of theories: conceptual tools developed by a research community to improve the problem-solving efficacy of its body of knowledge.
This is partly from a feeling that if we try to get in the business of line drawing “theory” vs “not theory”, when we really mean to make an assessment of a theory’s various qualities, we quickly run into trouble.
A relevant snapshot from a current manuscript in prep:
@dlevenstein
“Theory” feels a bit like “cause” insofar as people mean different things by it (and as you say, no line should be drawn about what qualifies). Descriptive words might help, but I also wonder if (unlike causality) it defies classification. It’s difficult for me to imagine a taxonomy for it.
@NicoleCRust theory sketch? In philosophy of mechanisms, a "mechanistic sketch" is used for incomplete mechanisms (though I think that's most theories / mechs..)
I guess all theories are sketches to some extent, the q is how detailed / precise it is, or if its qualitative / quantitative.
And I think Plato was famous for a 'path' or 'hierarchy' linking the 'perfect' case of any object or event, via various 'ideas' or 'views' of it as described or believed, to its phenomenon as observable 'on the ground'.
Would that system supply some useful definitions...?
IIUC you have three criteria: 1) development stage 2) explanatory power 3) peer assent/respect.
The word “Theory” is too grand but “hypothesis” too dismissive. You want intermediate commitment.
=> You have a “promising approach”, you’re recruiting for a “(proposed) research programme”, consisting of e.g. working hypotheses, a working model, a research concept? In math it’s a strategy but not a proof.
Alternatively it’s all theory/models and 1,2,3 are criteria/attitudes…
@NicoleCRust My own usage is to call Copernicus' contribution a theory. Whatever is confirmed by abductive inference/IBE, counts as a theory by me. Very low bar. (My guess is that I'm a little idiosyncratic about this.)
@NicoleCRust
It sounds like you are seeking "effective theory".
Effective theory is a theory that explains the observations or mechanism phenomena well, but is not considered fundamental. Classical theory of gravity is a great effective theory, but it's not fundamental theory (anymore).
In the same way the current field theories are considered "effective" theories at the low energy limits. There must be something else that explains all energy ranges.
@maxpool
Yes, I think that’s right - thanks! Although from the blog, I suspect that these theories might not get enough credit (in terms of the delta of their contribution to pushing things forward).
@NicoleCRust
I guess I like "idea" or "thoughts" just because people don't tie them to a specific stage of the process, and they don't sound "finished," whatever that is.
@NicoleCRust I use the word "model" a lot--it's more than a hypothesis, it should be a bit more substantive, it makes predictions and can be tested but it isn't quite as weighty as a theory.
@NicoleCRust@chiasm Given the examples, it's interesting alright. Copernicus and Kepler definitely had models (and perhaps theories too, that's the debate), but did Galileo have a model? He did have theories.
@billseitz
I have mixed thoughts. On one hand, we don’t want to impose something on the whole that only accounts for a slice, I guess. But we also need words to help is differentiate.
Add comment