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


twitter 2015
     facebook 2015

Program

TUESDAY, JULY 7  /  15:30 - 17:00  /  DS-M260
Organized session / standard talks
Modeling real patterns in biology
Organizer(s):

Collin Rice (Lycoming College, United States)

The participants of this session agree that abstract and idealized models in biology capture real patterns in the world. The participants also agree that these models can explain those real patterns. What is at issue, then, is the status of the patterns and, accordingly, the form of explanation. According to Rice, appreciating the explanatory role of these idealized models motivates a departure from a causal account of explanation. He develops an account of explanation according to which explanatory relevance amounts to contextually salient counterfactual relevance. Walsh, Ariew, and Matthen, drawing on their earlier work, argue that these evolutionary patterns—patterns in natural selection and drift—are statistical in nature. They contrast this view with a causal interpretation of these patterns, which they call the traditional view. Finally, Potochnik motivates a view of population biology models that she takes to be intermediate between causal and non-causal construals. In her view, these models represent actual causes, but they do so in a partial and skewed way, due to an emphasis on causal patterns instead of individual causal processes. Thus, in this session, we hope to clarify what is at stake among these alternative interpretations of the explanatory value of models in biology, and of the real patterns those models capture.


Causal patterns and actual causes

Angela Potochnik (University of Cincinnati, United States)

The use of optimality models and evolutionary game theory in population biology has been viewed by some as a commitment to adaptationism — to positing natural selection as the only influence, or the most important influence, on evolution. For most philosophers and many biologists, such a commitment is unwelcome. In contrast, some have recently motivated the idea that these modeling approaches succeed in virtue of their distance from representing actual causes at all. These views of how optimality and game theory models relate to biological systems are opposed. The first suggests that these models are intended to represent all causal influences on evolutionary outcomes, or at least all of the most important influences, while the second maintains that these models are not intended to accurately represent anything at all about the actual causal influences on those outcomes. In this talk I introduce an interpretation of the aim of these modeling approaches that is distinct from, and in some regards intermediate to, these two alternatives. In my view, optimality and game theory models are designed to provide a partial and skewed representation of the (actual) causal influences on evolutionary outcomes. The partial and skewed nature of the representation is due to a focus not on the causal influences in individual instances of evolution, but on causal patterns across systems. But this still serves to represent actual causes. I will motivate this interpretation of optimality models and evolutionary game theory by examining how the models tend to be employed, what work is typically done to justify their assumptions, and the direction this research appears to be taking. In my view, this case study illustrates a quite general feature of science: partial and skewed representation that facilitates focus on causal patterns.


Four pillars of statisticalism

André Ariew (University of Missouri, United States); Denis Walsh (University of Toronto, Canada); Mohan Matthen (University of Toronto, Canada)

An evolutionary population dynamics model explains the large-scale patterns of change and stasis in the trait structure of a population. They do so by describing the variation in fitness in the population. When a population varies in its fitness, it is said to be undergoing natural selection. When the outcome of a population differs from that predicted by the variation in fitness, the population is said to be undergoing drift. Over the past fifteen years there has been a considerable amount of debate about what theoretical population dynamic models tell us about biological reality. Two major positions have emerged, the traditional and the statistical. While the debate between the orthodox and statistical factions has been vigorous, it has not always been particularly productive or germane. This may be due, in large measure, to a widespread misapprehension of the statisticalist position. Our objective here is to outline as clearly and simply as possible the fundamental features of the statistical interpretation, in an attempt to forestall some of the more common misunderstandings. As we see it, statisticalism rests on four core commitments (i) evolution is a higher-order effect (as are selection and drift), (ii) trait fitness is a primitive concept in population models, (iii) population dynamics models explicitly represent only the relative growth rates of abstract trait types, and (iv) selection and drift are description dependent, that is to say a population can only be said to be undergoing selection and/or drift relative to a model. Together, these constitute the four pillars on which the statistical interpretation rests. Our objective in this presentation is to expand on these four core commitments and provide reasons for why we remain committed to statisticalism in the face of objections from the traditional perspective.


Biological patterns, idealization, and counterfactuals

Collin Rice (Lycoming College, United States)

Biological modelers are often interested in explaining patterns that hold across populations that are extremely heterogeneous. One prominent view of scientific explanation suggests that idealizations can contribute to explanations by distorting only those features that do not make a difference to the explanandum. However, many biological models explain patterns despite their distortion of the difference-making features of their target systems. Moreover, in many cases we are simply unable to “decompose” a scientific model into its accurate and inaccurate parts. That is, we are unable to identify which features are misrepresented and to what degree they are misrepresented. As a result, we need to expand our account of the positive contributions idealizations make to explanations of biological patterns. In this talk, I present an account of explanation that focuses on providing modal information about which contextually salient features are counterfactually relevant and irrelevant to the pattern of interest. On this account, the set of counterfactual information required to explain can be provided in numerous ways—e.g. by a causal model, a statistical model, or an equilibrium model. Moreover, I show that models can provide the counterfactual information required to explain even when they misrepresent difference-making features of their target systems. This account has several important virtues. First, the account allows us to see how idealized biological models are able to explain patterns even when we cannot decompose them into their accurate and inaccurate parts. Second, the account shows why providing explanations of patterns provides scientific understanding of why those patterns occur across a range of possible systems. Third, the account shows additional ways idealizations can contribute to the explanations of patterns even when they distort difference-making factors. As a result, I contend that we should adopt and continue to develop a counterfactual account of how idealized models explain biological patterns.