c·t = CONST, THAT CLICKS EPISODE 48 ／ Part VI — examining the procedure
"That is unnatural" was a declaration about the prior
Beauty was not
a third currency
Take it apart into six components and four reduce to brevity and fit.
What remained was a choice of prior — and one thing we could not measure.
Episode 5's balance measured two things: brevity (the price of a parameter) and fit (\(\Delta\chi^2\)). But physicists routinely invoke a third — "it is beautiful". This time we take it head-on: is beauty a third currency, or a restatement of the first two? And we write down honestly what we could not measure.
01Taking "beauty" apart
| Component | What it is | Which currency |
|---|---|---|
| Symmetry | reduces the independent parameters | reduces to brevity |
| Unification | reduces the number of inputs | reduces to brevity |
| Rigidity (no other choice available) | reduces free choices | reduces to brevity |
| Depth (hits what it was not built for) | fits data it was not fitted to | reduces to fit |
| Naturalness | a claim about the prior | does not reduce |
| Sensory pleasure | how the equation feels to look at | not measurable |
Conclusion of §01
Four of the six reduce to the two existing currencies.
What remains is naturalness (measured in §03) and sensory pleasure (§06).
02What does symmetry buy? — counting with the CKM matrix
a compression of 2.25×, worth 26.8 bits at Episode 5's price
Symmetry is measurable as brevity. There is no mystery here.
03The core — "naturalness" was a claim about the prior
| Quantity | With a linear prior | With a log-uniform prior | Difference |
|---|---|---|---|
| \(v/M_{\rm Planck}\) (the hierarchy problem) | \(55.5\) bit | \(6.3\) bit | \(49.1\) |
| \((v/M_{\rm Planck})^2\) (Higgs tuning) | \(110.9\) bit | \(6.3\) bit | \(104.6\) |
| \(\rho_\Lambda/\rho_{\rm Planck}\) (Ep. 32) | \(408.4\) bit | \(8.2\) bit | \(400.2\) |
The main point of this episode
The same number is either 55 bits surprising or 6 bits surprising.
Saying "that is unnatural" is the same as declaring that the prior is linear —
exactly the place Episode 19 §01 flagged as "depends on how the prior range is drawn".
── So: naturalness is not a third currency; it is a choice of prior.
Figure: the surprise attached to the same small number under two priors. A linear prior turns orders of magnitude straight into bits; a log-uniform prior only takes the logarithm of the number of decades. Move the slider — "unnaturalness" is decided by which ruler you apply.
04But sometimes a linear prior is justified
\(\theta\) is an angle, so there is a reason for a uniform prior on \([0,2\pi)\)
$$-\log_2\frac{10^{-10}}{2\pi}=\mathbf{35.9\ \text{bits}}$$Conclusion of §04
This is a genuine fine-tuning — there is no escaping into a log-uniform prior.
── When invoking "naturalness", ask whether there is a reason for the prior.
An angle has a reason. A ratio of masses, so far, does not.
05Taking famous "beautiful theories" apart
| Theory | What the beauty consists of | Currency | Measurable? |
|---|---|---|---|
| General relativity | rigidity (nearly unique given the assumptions) | brevity | yes |
| The Dirac equation | it predicted the positron (unbuilt-for) | fit | yes |
| The Standard Model | 32 parameters | not brief | not called beautiful |
| Supersymmetry | solves the hierarchy problem (naturalness) | the prior | depends on the prior |
| String theory | a uniqueness claim → the landscape | the brevity claim collapsed | unsettled |
The reasons a theory is called beautiful usually decompose into brevity or fit. When they do not, suspect the prior.
06What we could not measure
Conclusion of §06
Written down honestly: this cannot be measured here.
── Not measurable is not the same as not real. It simply falls outside this series' currencies.
And "do not use in a verdict what you cannot measure" has been the practice since Episode 3.
07Holding the series itself against this ruler
| Currency | For \(c\cdot t=\)const | Verdict |
|---|---|---|
| brevity | one fewer parameter (Ep. 25) | it buys |
| fit | \(q\) and \(w\) miss by orders (Ep. 46) | it pays heavily |
| naturalness | contains no unnatural numbers | a prior question — does not bite |
| sensory pleasure | the appeal of the word "constant" | not measurable |
It buys with brevity and pays heavily with fit — the verdict does not change. And "the shape of it is pleasing" does not enter the verdict — that was the procedure.
(1) §01's "six components" is this series' decomposition. Discussion of beauty in physics has a long history (Dirac, Weinberg, and more recently Hossenfelder's critical account), and there is no settled way to divide it up — the six here are a convenience, sorted by whether this tool can handle them.
(2) The 80 decades behind §03's "6.3 bits log-uniform" are arbitrary. The value moves with where the prior range is drawn — the point is the structure "linear versus log-uniform moves it by 50 bits", not the number 6.3 (the same caution as Episode 19 §01).
(3) §04's "an angle justifies a linear prior" is not absolute either. Depending on how a high-energy theory generates \(\theta\), non-uniform priors are arguable (in axion models \(\theta\) relaxes dynamically) — read it as "a linear prior is easier to justify here".
(4) §05's assessments are summaries. "General relativity is nearly unique" is a statement in the sense of Lovelock's theorem and depends on the assumptions; "string theory's uniqueness claim collapsed" is a point on which views differ depending on how one assesses the landscape — this document endorses none of these theories.
(5) §06's "cannot be measured" is a statement about this series' tools. It is not a claim that aesthetic judgement cannot be treated in any framework — cognitive science and philosophy of science have their own approaches. All that is claimed here is that it is none of description length, fit, or prior.
Exercises
- How many of the six components of "beauty" reduce to the two existing currencies?
Show the answer
Four — symmetry, unification and rigidity reduce to brevity; depth (hitting what it was not built for) reduces to fit. What remains is naturalness (a prior question) and sensory pleasure (not measurable). - How much does symmetry buy in the CKM matrix?
Show the answer
Nine parameters of a 3×3 unitary matrix, minus five removed by rephasing, leaves four (3 angles + 1 phase) — worth 26.8 bits at Episode 5's price. Symmetry is measurable as brevity, and there is no mystery here. - How surprising is \(\rho_\Lambda/\rho_{\rm Planck}=1.13\times10^{-123}\)?
Show the answer
It depends on the prior — 408.4 bits under a linear prior (Episode 32's number), 8.2 bits under a log-uniform one over 300 decades. A difference of 400 bits. Saying "that is unnatural" is the same as declaring a linear prior. - Why is the strong CP problem different from the two in §03?
Show the answer
Because \(\theta\) is an angle — a uniform prior on \([0,2\pi)\) has a reason, and there is no escaping into a log-uniform prior. So 35.9 bits is a genuine fine-tuning. When invoking naturalness, ask whether there is a reason for the prior. - (Harder) What could this series' tools not measure?
Show the answer
Sensory pleasure — not description length, not fit, not the prior. Not measurable is not the same as not real, but it falls outside these currencies — and "do not use in a verdict what you cannot measure" has been the practice since Episode 3.
Summary: beauty was not a third currency
Taking "beauty" apart into six components, four reduce to the two existing currencies — symmetry, unification and rigidity to brevity, depth to fit. And what symmetry buys can actually be counted: the CKM matrix's nine parameters become four under rephasing, a saving of 26.8 bits.
What remained was naturalness — and it turned out not to be a third currency but a choice of prior. \(v/M_P=2\times10^{-17}\) is 55.5 bits surprising under a linear prior and 6.3 under a log-uniform one; \(\rho_\Lambda/\rho_{\rm Planck}\) is 408.4 bits versus 8.2, a difference of 400. Saying "that is unnatural" is the same as declaring the prior to be linear — exactly the place Episode 19 §01 flagged from the start.
But a linear prior is sometimes justified — the strong CP problem's \(\theta_{\rm QCD}<10^{-10}\) concerns an angle, so a uniform prior on \([0,2\pi)\) has a reason and there is no escape. 35.9 bits, a genuine fine-tuning. When invoking naturalness, ask whether there is a reason for the prior — an angle has one; a ratio of masses, so far, does not.
And there was something we could not measure — sensory pleasure. Not description length, not fit, not the prior. Written down honestly: this cannot be measured here. Not measurable is not the same as not real, but it falls outside these currencies — and "do not use in a verdict what you cannot measure" has been the practice since Episode 3.
Finally we held the series against its own ruler. \(c\cdot t=\)const buys with brevity and pays heavily with fit — the verdict does not change. The pleasing shape of the word "constant" does not enter the verdict.
This document is Episode 48 of "c·t = const, That Clicks" (the third of Part VI), written for physics-minded high-school and university readers. Minimum description length, the relation between naturalness and priors, and the strong CP problem are all standard, and nothing here is a new claim — the numbers are computed in kenshou/calc52.py. §01's "six components" is this series' decomposition; discussion of beauty in physics has a long history (Dirac, Weinberg, and more recently Hossenfelder's critical account) and there is no settled way to divide it up — these six are a convenience, sorted by what this tool can handle. The 80 decades behind §03's log-uniform figure are arbitrary and the value moves with the prior range — the point is that linear versus log-uniform moves it by 50 bits, not the number itself. §04's "an angle justifies a linear prior" is not absolute: depending on how a high-energy theory generates \(\theta\), non-uniform priors are arguable (in axion models \(\theta\) relaxes dynamically). §05's assessments are summaries — "general relativity is nearly unique" is meant in the sense of Lovelock's theorem and depends on the assumptions, and how one assesses the string landscape is a matter on which views differ; this document endorses none of these theories. §06's "cannot be measured" is a statement about this series' tools, not a claim that aesthetic judgement is beyond every framework. ── To make a PDF, use your browser's Print dialogue (sliders freeze and answers are hidden in the print version).