fitness and prediction in evolutionary theory
TRANSCRIPT
Fitness and Prediction
in Evolutionary Theory
London School of Economics, 25/1/2017
Charles H. Pence
Department of Philosophy
Outline
1. Fundamentals of Selection2. Two Roles for Fitness3. e Claim4. Predictive Fitness
4.1 From the PIF to Predictive Fitness4.2 Ways Out?
5. e Moral
e take-home: A common conceptual analysis of tnessin the philosophy of biology is mistaken.
Charles H. Pence 2 / 53
Natural Selection
Charles H. Pence Fundamentals of Selection 3 / 53
t = 0
t = 1 t = 2
Charles H. Pence Fundamentals of Selection 4 / 53
t = 0 t = 1
t = 2
Charles H. Pence Fundamentals of Selection 4 / 53
t = 0 t = 1 t = 2
Charles H. Pence Fundamentals of Selection 4 / 53
Fitness
Charles H. Pence Fundamentals of Selection 5 / 53
>
Charles H. Pence Fundamentals of Selection 6 / 53
>
Charles H. Pence Fundamentals of Selection 6 / 53
>
Charles H. Pence Fundamentals of Selection 6 / 53
Blue organisms leave more ospringthan red organisms.
A circle: the tautology problem
Charles H. Pence Fundamentals of Selection 7 / 53
Blue organisms leave more ospringthan red organisms.
A circle: the tautology problem
Charles H. Pence Fundamentals of Selection 7 / 53
Blue organisms will probably (aredisposed to) leave more ospring than
red organisms.
Charles H. Pence Fundamentals of Selection 8 / 53
The Propensity
Interpretation of
Fitness
Charles H. Pence Fundamentals of Selection 9 / 53
Charles H. Pence Fundamentals of Selection 10 / 53
Two Notions of
Fitness
Charles H. Pence Two Roles for Fitness 11 / 53
Matthen and Ariew (2002)
[F]or many this notion of an organism’s overall competitiveadvantage traceable to heritable traits is at the heart of thetheory of natural selection. Recognizing this, we shall callthis measure of an organism’s selective advantage itsvernacular tness. According to one standard way ofunderstanding natural selection, vernacular tness – orrather the variation thereof – is a cause of evolutionarychange. (p. 56)
Charles H. Pence Two Roles for Fitness 12 / 53
Matthen and Ariew (2002)
Fitness occurs also in equations of population geneticswhich predict, with some level of probability, thefrequency with which a gene occurs in a population ingeneration n + 1 given its frequency in generation n. Inpopulation genetics, predictive tness (as we shall call it) isa statistical measure of evolutionary change, the expectedrate of increase (normalized relative to others) of a gene . . .in future generations. . . . (p. 56)
Charles H. Pence Two Roles for Fitness 13 / 53
Causal (vernacular) tness: general(causal) notion in natural selection
Predictive (mathematical) tness:predict future representation fromcentral tendency/expected value
Charles H. Pence Two Roles for Fitness 14 / 53
Matthen and Ariew (2002)
[N]atural selection is not a process driven byvarious evolutionary factors taken as forces;rather, it is a statistical “trend” with these factors(vernacular tness excluded) as predictors.ese theses demand a radical revision ofreceived conceptions of causal relations inevolution. (p. 57)
Charles H. Pence Two Roles for Fitness 15 / 53
The Claim
Charles H. Pence The Claim 16 / 53
ContraMatthen and Ariew, predictivetness is not a fruitful way tounderstand tness in biology.
By extension, neither is the dichotomybetween causal and predictive tness.
Charles H. Pence The Claim 17 / 53
It fails in a way that should lead us toquestion not only the predictive-causaldichotomy, but the relationship betweenshort-term and long-term evolutionary
processes.
Charles H. Pence The Claim 18 / 53
Predictive Fitness
Charles H. Pence Predictive Fitness 19 / 53
Standard view: predictive tness trackssomething like a central tendencyextracted from the distribution of
outcomes of interest
Charles H. Pence Predictive Fitness 20 / 53
Matthen and Ariew (2002)
e basic principle of statistical thermodynamics is thatless probable thermodynamic states give way in time tomore probable ones, simply by the underlying entitiesparticipating in fundamental processes. [. . . ]e same istrue of evolution. (p. 80)
Charles H. Pence Predictive Fitness 21 / 53
Lewontin (1974)
When we say that we have an evolutionary perspective ona system or that we are interested in the evolutionarydynamics of some phenomenon, we mean that we areinterested in the change of state of some universe in time.[. . . ]e sucient set of state variables for describing anevolutionary process within a population must include someinformation about the statistical distribution of genotypicfrequencies. (pp. 6, 16; orig. emph.)
Charles H. Pence Predictive Fitness 22 / 53
Among other reasons, this is to be usefulfor grounding medium- to long-termpredictions about evolutionary change
in tness comparisons.
Charles H. Pence Predictive Fitness 23 / 53
Charles H. Pence Predictive Fitness 24 / 53
Charles H. Pence Predictive Fitness 24 / 53
Fitness
Maximization
Charles H. Pence Predictive Fitness 25 / 53
One way to get these predictions:
Evolution as a process that maximizestness over time
Charles H. Pence Predictive Fitness 26 / 53
Birch (2016)
Neither [of the approaches surveyed] establishes amaximization principle with biological meaning, and Iconclude that it may be a mistake to look for universalmaximization principles justied by theory alone. (p. 714)
Charles H. Pence Predictive Fitness 27 / 53
But this isn’t the only way to generatepredictions. What about buildingpredictions from the PIF itself?
Charles H. Pence Predictive Fitness 28 / 53
From Causal to
Predictive
Charles H. Pence From the PIF to Predictive Fitness 29 / 53
How should defenders of the PIFrespond to M&A’s emphasis on
predictive tness?
Charles H. Pence From the PIF to Predictive Fitness 30 / 53
Millstein (2016)
Fitness is an organism’s propensity to survive andreproduce (based on its heritable physical traits) in aparticular environment and a particular population over aspecied number of generations.at is what tness is.[. . . ] Expected reproductive success is not the propensityinterpretation of tness and it never was; it was just oneway of trying to grapple with probability distributions andassign tness values. (p. 615)
Charles H. Pence From the PIF to Predictive Fitness 31 / 53
Charles H. Pence From the PIF to Predictive Fitness 32 / 53
e basic idea: Dene the propensityinterpretation in terms of facts about thepossible lives an organism (with a givengenotype, in a given environment) could
have lived.
Charles H. Pence From the PIF to Predictive Fitness 33 / 53
F(G , E) = exp( limt→∞
1t ∫ω∈Ω Pr(ω) ⋅ ln(ϕ(ω, t)) dω)
Having our cake and eating it, too:
• Gives you a mathematical model. . .• . . . that’s grounded in a specic dispositionalproperty
Charles H. Pence From the PIF to Predictive Fitness 34 / 53
Can we draw any conclusions about thequality of predictions from the Pence &
Ramsey model?
Charles H. Pence From the PIF to Predictive Fitness 35 / 53
F(G , E) = exp( limt→∞
1t ∫ω∈Ω Pr(ω) ⋅ ln(ϕ(ω, t)) dω)
A long-run limit: perfect for prediction! But itrelies on an assumption: non-chaotic populationdynamics
Charles H. Pence From the PIF to Predictive Fitness 36 / 53
Question: How common is non-chaoticdynamics in evolving systems?
My assumption: Won’t be able to answerthis – model is far too general.
Charles H. Pence From the PIF to Predictive Fitness 37 / 53
Approach of Doebeli & Ispolatov (2014):Investigate by simulating populations
with two features:
1. Density-dependent selectionpressures
2. High-dimensional phenotype space
Charles H. Pence From the PIF to Predictive Fitness 38 / 53
Doebeli & Ispolatov (2014)
Our main result is that the probability of chaos increaseswith the dimensionality d of the evolving system,approaching 1 for d ∼ 75. Moreover, our simulationsindicate that already for d ≳ 15, the majority of chaotictrajectories essentially ll out the available phenotypespace over evolutionary time. . . . (p. 1368)
Charles H. Pence From the PIF to Predictive Fitness 39 / 53
-100 0 100
-100
0
100
x1
x2
Doebeli and Ispolatov (2014), g. 5
Charles H. Pence From the PIF to Predictive Fitness 40 / 53
What’s the real-world timescale here?
How does it relate to the rate ofenvironmental change?
Charles H. Pence From the PIF to Predictive Fitness 41 / 53
A surprising result at this level ofgenerality. In cases where it holds, whatkinds of tness-based predictions would
remain viable?
Charles H. Pence From the PIF to Predictive Fitness 42 / 53
Charles H. Pence Ways Out? 43 / 53
Grant (1997)
e invasion is exponential, but nonlinear dynamics of theresident type produce uctuations around this trend.[Fitness] can therefore be most accurately estimated by theslope of the least squares regression of [daughterpopulation size] on t. (p. 305)
Charles H. Pence Ways Out? 44 / 53
Two major problems:• Loses sight of the dynamics• Oers poor predictions
Charles H. Pence Ways Out? 45 / 53
Move the goalposts:• Abandon all but short-term predictions• Abandon predictive content; predict simply‘chaos’
• Abandon prediction, focus on statisticalretrodiction
Charles H. Pence Ways Out? 46 / 53
• Resuscitate central tendencies: poorpredictions that ignore (interesting)dynamics
• Move the goalposts: loses themedium-to-long-term predictions citedabove
Charles H. Pence Ways Out? 47 / 53
The Moral
Charles H. Pence The Moral 48 / 53
Many uses of tness:
• Mathematical parameter in models• Causal property• Proxies for strength of selection inpopulations
• Statistical estimator for any of the above
Charles H. Pence The Moral 49 / 53
Fitness concepts are far more complexthan a dichotomy between two simple
roles for tness.
Charles H. Pence The Moral 50 / 53
What’s the relationship betweenshort-term and long-term evolutionary
processes?
Charles H. Pence The Moral 51 / 53
Complete continuity, all the way to thelong run? Unlikely.
Medium-term continuity?reatenedby these results!
Charles H. Pence The Moral 52 / 53
Complete continuity, all the way to thelong run? Unlikely.
Medium-term continuity?reatenedby these results!
Charles H. Pence The Moral 52 / 53
F(G , E) = exp( limt→∞
1t ∫ω∈Ω Pr(ω) ⋅ ln(ϕ(ω, t)) dω)
• Multi-generational life histories• Changing genotypes and environments overtime
• Disposition (propensity) dened over modalfacts about other possible lives of organisms
Charles H. Pence Bonus! 1 / 4
F(G , E) = exp( limt→∞
1t ∫ω∈Ω Pr(ω) ⋅ ln(ϕ(ω, t)) dω)
• Multi-generational life histories
• Changing genotypes and environments overtime
• Disposition (propensity) dened over modalfacts about other possible lives of organisms
Charles H. Pence Bonus! 1 / 4
F(G , E) = exp( limt→∞
1t ∫ω∈Ω Pr(ω) ⋅ ln(ϕ(ω, t)) dω)
• Multi-generational life histories• Changing genotypes and environments overtime
• Disposition (propensity) dened over modalfacts about other possible lives of organisms
Charles H. Pence Bonus! 1 / 4
F(G , E) = exp( limt→∞
1t ∫ω∈Ω Pr(ω) ⋅ ln(ϕ(ω, t)) dω)
• Multi-generational life histories• Changing genotypes and environments overtime
• Disposition (propensity) dened over modalfacts about other possible lives of organisms
Charles H. Pence Bonus! 1 / 4
State space of possible lives: Ω ∶ f ∶ R→ Rn
Cardinality, though, is too big! N(Ω) = 22ℵ0
Can still dene a σ-algebra F and a probability measurePr if we restrict our attention to:
• continuous functions ω,• functions ω with only point discontinuities, or• functions ω with only jump discontinuities
(Nelson 1959)
Charles H. Pence Bonus! 2 / 4
With that state space dened, we need three simplifyingassumptions:
1. Non-chaotic population dynamics2. Probabilities generated by a stationary randomprocess
3. Logarithmic moment of vital rates is boundede last two are trivial in our context.
Charles H. Pence Bonus! 3 / 4
en the following limit exists:
a = limt→∞
1tEw ln(ϕ(t)),
with Ew an expectation value (Tuljapurkar 1989). Fitness isthe exponential of this quantity a, and is equivalent (underfurther simplifying assumptions) to net reproductive rateand related to the Malthusian parameter.
Charles H. Pence Bonus! 4 / 4