MENTAL MODEL · M9818

Solomonoff Induction

TechnicalHigh supportMathematics

Included

Updated 2026-08-15

Loading revision record…

INTRODUCTION

English translation pending.

CORE DEFINITION

Solomonoff induction, developed by Ray Solomonoff in the 1960s, is a mathematical formalisation of the idea that simpler explanations deserve higher prior probability. A hypothesis is identified with a computer program that could generate the observed data, and hypotheses are weighted by two raised to the power of minus the length of the shortest such program, so short programs dominate. The core proposition is that Occam's razor becomes a theorem rather than a taste, and Bayesian updating then gives an ideal but uncomputable universal predictor. The key qualification is that the scheme is an unreachable limit: the halting problem makes exact shortest-program length uncomputable, so it functions as a principle, not an algorithm.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

- Prefer The Shorter: give more initial credence to the explanation with the shortest generating description. - Weight All Programs: remember every consistent hypothesis gets some weight, none is ruled out a priori. - Use As Principle: treat simplicity as a tie-breaker in reasoning, never as a computable score.

anchor

Anchor fast decisions

There are always infinitely many programs consistent with the data so far, so raw observation cannot select among them. Short programs are exponentially more numerous in effective weight than long ones, because there are far fewer short descriptions, which makes simple hypotheses jointly dominate the prediction. As new data arrives, Bayesian updating multiplies each program's weight by how well it predicts, and the surviving mass concentrates on the simplest adequate descriptions, which is the behaviour Occam's razor recommends.

MINIMUM ACTION

In progress 0/1

Practice this model in one real situation:

List the competing explanations for the observation in front of you. 2. Estimate the descriptive complexity of each, as though specifying it to a machine. 3. Give the shorter explanations the higher prior plausibility. 4. Update with each new piece of evidence rather than locking in a favourite. 5. Remember the scheme is a guiding ideal, and never present the ranking as a computed number.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
en.wikipedia.orghttps://en.wikipedia.org/wiki/Solomonoff%27s_theory_of_inductive_inferenceZH · Explicit
verified

RELATED MODELS