International Society for History, Philosophy, and Social Studies of Biology

twitter 2015
     facebook 2015


MONDAY, JULY 6  /  09:00 - 10:30  /  DS-R515
Organized session / standard talks
Probability and chance in biology

Joel Velasco (Texas Tech University, United States)

Statements about probability and chance occur all throughout the biological sciences and philosophers have long been interested in how to interpret such statements. As our empirical knowledge grows, the complexity of our models grow and philosophical understanding becomes more difficult and more important at the same time. Sometimes, these questions are approached by detailed examination of the science itself, other times with a focus on general considerations about the nature of probability. Often both. In this symposium, the three presenters will interact by approaching similar questions from different places.

First, Charles Pence will discuss recent work in biology on adaptive population dynamics and how these result fit within his account of fitness (co-authored with Grant Ramsey). Arguing against any conception of individual fitness, Marshall Abrams will defend his own account of type fitness by defining it in terms of a particular interpretation of probability, a ‘complex causal structure’ interpretation. Joel Velasco will then step back and argue that we should not be thinking about probabilities in biology generically, but rather, that in different sub-disciplines and in different contexts even within the same problem, the role of probability statements is quite different.

Adaptive dynamics, chaos, chance, and fitness

Charles Pence (Louisiana State University, United States)

In recent work, Grant Ramsey and I (2013) defended a picture of individual fitness which, while drawing on the same intuitions as the propensity interpretation of fitness (PIF), manages to resolve many of the counterexamples to that model which philosophers have offered in the intervening thirty years since its proposal. A crucial role in that argument was played by results from mathematical population modeling, particularly from a research tradition known as adaptive dynamics. While adaptive dynamics offered the mathematical resources necessary to resolve several problems with the PIF, it also offers more than this – in particular, it provides a broader formalism which we can use to think about the trajectory of evolving populations over time. In another recent paper, Michael Doebeli and Iaroslav Ispolatov (2014) demonstrate a surprising result: using models very similar to ours, they derive that, as the dimensionality of the evolving system increases, the probability of chaotic behavior goes very rapidly to 1. They argue that on the basis of these results, we should doubt that the current phenotypic state of a population can ever “be understood as the result of an equilibrium or optimisation process” (2014, p.1369). In this talk, I will explore the intersection of these two sets of results. If adaptive dynamics does offer us a fruitful way to understand individual fitness, what is the impact on that model of these results concerning long-term chaos? What does it say about our understanding of the sense in which individual fitness, and by extension, natural selection, is probabilistic? Do these results contradict our earlier claim that adaptive dynamics can provide stable values for individual fitnesses? Things are not, I will argue, that dire. But they do necessitate a rethink in the conceptual role of fitness and its close cousin, the principle of natural selection.

Resolving puzzles about types and tokens in evolution with new probability concepts

Marshall Abrams (University of Alabama at Birmingham, United States)

Philosophers concerned with whether natural selection is a cause typically assume that if fitness differences are in some sense causal, the must be defined as "token fitnesses" in terms of propensities of outcomes for token organisms (lineages, etc.) in particular circumstances. On the view that natural selection concerns heritable biological types (Lewontin 1970; Godfrey-Smith 2009), advocates of this view typically assume that fitnesses of types ("type fitnesses") are averages of token fitnesses of those organisms actually in a population. That this is the received view despite significant, unaddressed challenges known for several years (Abrams 2006, 2012, 2014; Ariew & Ernst 2009; Sober 1984, 2001, 2013) may be due the lack of an alternative, causal conception of type fitness. The goal of this talk is to describe such an alternative--one that makes type fitnesses fundamental, and causal--by defining fitnesses in terms of a "complex causal structure" (CCS) interpretation of probability (Rosenthal 2010, 2012; Strevens 2011; Abrams 2012a,b). I treat recurrent realizations of heritable biological traits in a population and environment as defining a kind of causal process or device, with states of organisms and their environment as inputs, and instances of outcomes relevant to evolution as outputs. Such a device can realize CCS probabilities if it has the kind of internal structure underlying Strevens' "microconstancy" or my "bubbliness". I argue that biological populations are likely to generate such causal structure, and that other conditions for my "FFF mechanistic probability" are likely to be satisfied. Importantly, fitnesses defined in terms of these probabilities, though causal, attach to types rather than tokens or particular trials (Rosenthal 2010, 2012; Abrams 2012a,b). My strategy doesn't require biologically questionable assumptions about physical properties, as Strevens' (2003, 2008, 2011, 2013) discussions of evolutionary processes do.

A disunified account of objective probability in biology

Joel Velasco (Texas Tech University, United States)

I will be examining the use of probability as it appears in three different subdisciplines in biology. First, within phylogenetics where we use and infer molecular clocks and various rates of evolution, as well as make pronouncements about the likely age of various clades or fossils. Second, in ecology and evolution where models of foraging behavior and genetic drift are used to explain the distribution of traits in species. And third, in molecular biology where stochastic models of molecular movements and processes are used to explain stochastic gene expression and “noisy” development. Questions of interpretation have arisen in each of these cases and typically, questions focus on whether such probabilities must be subjective or can be understood in an objective way. I will argue that while some useful distinctions can be made, classifying any of these as "subjective" or "objective" can be misleading and is mostly unhelpful. While all three cases are different, each has some elements (different between the three) that are typical for "objective" claims about the world and all share some aspects typical of "subjective" (or much better, "epistemic") interpretations. I will then use these cases as the basis for an argument that we are unlikely to be able to have a unified account of probability even in relatively narrow domains such as “explanatory models in biology”, much less a unified account across all of the sciences.