Variance Bounds Selection Response

Prime #

1264

Origin domain

Mathematics Logic

Subdomain

evolutionary dynamics → Mathematics Logic

Core Idea

The rate at which a sorting process shifts a population's mean along a trait equals — or is bounded by — the within-population variance of the selection-relevant quantity, weighted by selection intensity (the Price equation, with Fisher's theorem and the breeder's equation as specialisations). Selection consumes the variance it acts on, so persistent adaptive change requires a variance-regenerating mechanism: variance is the fuel, sorting the engine.

How would you explain it like I'm…

 

No faithful explanation at this level. All three generators marked eli5 na: any 5-year-old framing collapses into the misconception 'picking/pushing the best harder makes things change faster,' which erases the prime's load-bearing content — that the speed is set (and bounded) by the population's existing variance, so with zero spread there is no response at any selection strength, and that selection consumes the very variance it needs.

Variety Sets the Speed

Variance Bounds Selection Response is about how fast a group changes when something keeps picking winners. Picture a class running a race where the fastest kids get copied into the next class. How quickly the average speed goes up depends on two things together: how SPREAD OUT the speeds already are (variance), and how strongly you favor the fast ones (selection intensity). The big surprise is that if everyone is exactly the same speed, no spread, then no amount of favoring the fast ones changes the average at all. And because picking winners uses up the spread (soon everyone is fast), you need something to keep making new variety, or the change stops.

Change ≈ Variance × Pressure

Variance Bounds Selection Response says the *rate* at which a sorting process shifts a population's average on some trait equals — or is capped by — the *variance* of the selection-relevant quantity, times the *selection intensity*. The compact form is *rate-of-mean-change ≈ variance × selection-intensity*. So adaptation speed is governed by spread, not by pressure alone: no variance means no response at *any* selection strength, while high variance gives a big response under even modest pressure. The four pieces are a population of units each bearing a trait value, a sorting process that preferentially propagates units by that trait, the identity itself (the Price equation is the most general form, with Fisher's theorem and the breeder's equation as special cases), and a consumption consequence: selection acts on variance and thereby *consumes* it, collapsing variance as high-trait units dominate, so lasting change requires a variance-regenerating mechanism — mutation, recombination, exploration noise, innovation. Its content over a vague appeal to 'diversity' is the linear identity plus the consumption dynamic.

 

Variance Bounds Selection Response is the structural pattern in which the *rate* at which a sorting process shifts the mean of a population along some measurable trait equals — or is bounded by — the *within-population variance* of the selection-relevant quantity, weighted by the *selection intensity*. It rests on four commitments. There is a population of units — organisms, strategies, beliefs, cultural variants, portfolio assets, sampled candidates, model parameters — each bearing a value of some measurable trait. There is a sorting or selection process — differential reproduction, survival, copying, replication, weighting, sampling, gradient update — that preferentially propagates units according to that trait. There is an identity or inequality of the form *rate-of-mean-change ≈ variance × selection-intensity*, captured most generally by the Price equation, with Fisher's fundamental theorem and the breeder's equation as canonical specializations. And there is a variance-consumption consequence: selection, by acting on variance, also *consumes* it, as high-trait units come to dominate and variance collapses, so persistent adaptive change requires a *variance-regenerating mechanism* — mutation, recombination, exploration noise, innovation, rebalancing — to keep the loop fed. The diagnostic payoff is sharp: any question of the form 'how fast will this system adapt under this selection pressure?' is answered by asking 'what is the variance of the relevant trait, and what regenerates it?' rather than by asking about the selection pressure alone. No variance, no response — at any selection strength; high variance, large response — at modest pressure. Its distinctive content over a generic appeal to 'diversity' is twofold: the *identity* that quantifies the response as linear in variance, and the *consumption dynamic* that makes regeneration a structural necessity rather than an optional extra.

Broad Use

• Population genetics: Fisher's theorem (rate of mean fitness = additive genetic variance) and the breeder's equation \(R = h^2 S\).
• Evolutionary game theory: under replicator dynamics, the rate of change of mean payoff equals payoff variance among strategies.
• Cultural evolution: the rate of mean trait change under biased transmission is variance among bearers times the bias.
• Reinforcement learning / bandits: reward-improvement rate scales with reward-estimate variance; exploration is variance regeneration.
• Finance: Kelly-style rebalancing ties growth-rate improvement to variance across asset returns.
• Optimization / biotech: CMA-ES ties convergence to sampled-candidate covariance; directed evolution ties protein improvement to library diversity.
• Experimentation: A/B-testing improvement rate is set by variant-population variance, so low-variance programmes plateau.

Clarity

Dissolves the confusion between selection pressure and adaptive rate: a breeding programme without genetic diversity will not respond to stricter culls, making "which term is binding?" a checkable question and exposing intensity-side fixes as wasted when variance is the constraint.

Manages Complexity

Supplies a worklist — trait, variance, intensity, current rate, regeneration rate, binding term — that turns "our improvement has stalled" into a localisable diagnosis: out of variance, or selection too weak, with entirely different fixes.

Abstract Reasoning

Enables a precise counterfactual — holding intensity fixed, what happens to the rate if variance is doubled, halved, or zeroed? — answered linearly in variance, and lets the qualitative intuition travel without the Price-equation formalism until precision is needed.

Knowledge Transfer

• Breeding ↔ ML ↔ finance: low genetic diversity, exploration collapse, and a volatility floor are one problem with one borrowed fix (import fresh variance).
• The plateau diagnosis: a stalled A/B programme, an inbred line, and a converged RL policy all read as consumed-but-unregenerated variance.
• The exact carry: because the relation is an identity, the diagnosis and remedy move across systems without loss.

Example

A mature A/B-testing programme improves conversion 3% per quarter for eight quarters, then decays to half a percent at unchanged variant volume and rigour: selection has consumed the variance as designs converged, so the fix is variance regeneration (diversity quotas, exploration rounds), not sharper selection.

Relationships to Other Primes

Parents (1) — more general patterns this builds on

• Variance Bounds Selection Response presupposes, typical Variation Strategies — The file: this prime supplies the THEORETICAL NECESSITY that variation_strategies serve — variance-regeneration is mandatory because selection consumes the variance it acts on. Presupposes variation as the diversity-generating tactic family it quantifies.

Path to root: Variance Bounds Selection Response → Variation Strategies → Learning → Adaptation

Not to Be Confused With

• Variance Bounds Selection Response is not Selection Bias because selection bias is a measurement distortion from a non-representative sample, whereas this is a dynamical identity about how fast selection shifts a mean — same word, different kind of object.
• Variance Bounds Selection Response is not Variation Strategies because variation strategies are tactics for generating diversity, whereas this is the identity that makes them structurally necessary and quantifies response as linear in variance.
• Variance Bounds Selection Response is not Diversity as a virtue because diversity is a static stock, whereas this adds the consumption dynamic that makes diversity a flow to be regenerated, not a stock to be admired.