Occam's Razor and Parsimony
Among explanations that fit equally well, prefer the simpler; added complexity must earn its place by explaining more.
The Principle
Occam's razor advises that, among competing explanations that account for the evidence equally well, the one with fewer assumptions should be preferred. It is not a claim that nature is simple, but a rule for managing our own tendency to over-explain: do not multiply entities beyond necessity. Complexity is a cost to be justified, not a virtue.
Why Parsimony Works
A more complex model can fit any dataset better simply by having more freedom, including fitting the noise. Extra parameters buy in-sample fit at the risk of out-of-sample failure. Preferring the simpler model that fits adequately guards against mistaking flexibility for understanding, and it tends to generalize better to new data.
Formal Echoes
- Model-selection criteria penalize the number of parameters against goodness of fit.
- Regularization shrinks a model toward simplicity unless the data demand complexity.
- Bayesian inference favors models that make sharp predictions over vague, flexible ones.
- Description-length views equate the best model with the shortest faithful account of the data.
Not a License for Oversimplification
The razor cuts only between explanations that fit equally well. A simpler model that fails to account for the evidence is not preferred; simplicity never overrides fidelity. Einstein's counsel captures the balance: make things as simple as possible, but no simpler. The right complexity is the least that the phenomenon actually requires.
In Modeling Practice
Adding physics to a model must be justified by the phenomena it captures, not added for appearance. Kronos design work states which effects a model includes and which it omits, so added complexity is a deliberate, defensible choice rather than decoration, and a simpler model is used wherever it suffices.