The efficiency of information gain by natural selection.

Regret-based learning theory shows that selection’s inefficiency, measured via relative lineage fitness, can be bounded and experimentally validated in Escherichia coli, resolving inconsistencies with Kimura’s substitution-load model.

• Why it matters: Kimura’s relationship between information gain and substitution load is influential but fails to match selection experiments, leaving uncertainty about how efficiently evolving lineages learn environmental conditions.
• What they did: The authors model selection as a learning process and use regret to quantify inefficiency, deriving bounds for frequency-dependent selection and arbitrary environmental change, then testing them experimentally in Escherichia coli.
• The result: The extended framework recovers Kimura’s core insight and shows selection is asymptotically optimal for the basic learning problem, acquiring information more efficiently than comparable processes in complex environments.

AI summary of the abstract · may contain errors · not medical advice · read the paper.