Dissertation Abstract

Sarah M. Roe sroe@ucdavis.edu Department of Philosophy, University of California, Davis Dissertation Abstract The Salmon-Roe Approach to Mechanistic Explanation My dissertation is concerned with mechanistic approaches to explanation, particularly within the biological sciences. I argue that contemporary mechanistic approaches all suffer from several shortcomings, most notably issues regarding generalities, mechanism demarcation, negative causation, reduction, regularity, and scientific practice. In order to illustrate the severity of these problems, I offer an original biological case study regarding porcine lactation and breast health, drawing from my own scientific research endeavors, to illuminate key points throughout my dissertation. As a result, I argue that there is indeed reason to look elsewhere for a different mechanistic approach. I offer one such approach and argue that the Salmon-Roe approach to mechanistic explanations is capable of circumventing these difficulties. For the past three decades, the concept of ‘mechanism’ has regained centrality in science as well as philosophy. Recently, philosophers of science have become interested in mechanistic approaches to scientific explanation. Within the mechanism literature, one important group can be distinguished, the complex-systems approach, or the new mechanists, which consists of work by Machamer, Darden, Craver, Glennan, Bechtel and Abrahamsen. Instead of appealing to laws, the new mechanists argue that scientists are able to provide successful explanations by identifying and manipulating entities and activities in a model of a mechanism, thereby determining how they are situated in the mechanism. The explanation consists of explaining how those entities and activities act and interact to produce the phenomenon of interest. When all of the entities and activities are identified and organized correctly, a mechanism of interest will reliably and regularly produce the phenomenon to be explained. For example, the mechanism behind protein folding is often modeled through computer simulation. A computer program is created with the relevant entities and activities. For instance, the amino acid sequence of a particular protein of interest, important components found within a particular chemical environment, and the binding and twisting of certain entities are delineated. The simulation is then run to completion in order to recreate, and in this case slow down, the process of protein folding in an attempt to better understand how protein folding occurs given the variables and their organization. Contra the new mechanistic philosophy of science, I argue that, surprisingly, Wesley Salmon’s 1984 Mark Transmission account, thought by some to be irrelevant for the biological sciences, in fact provides a more comprehensive approach to mechanistic explanations within the biological sciences. My dissertation focuses on an often overlooked portion of Salmon’s work that combines the idea of causal processes with mark transmission, as a two tiered approach to scientific explanations (see Salmon 1984). Although Salmon does not provide an explicit definition of mechanisms, I extend Salmon’s account to incorporate mechanistic modeling. Causal processes and the intersections of those processes are the mechanisms that operate in our universe and are precisely what some scientists trace and model during their research in an attempt to explain phenomena. In addition to this, indirect or common cause information is gathered by scientists via the use of statistical relevance relations or the assembling of all facts statistically relevant to the phenomenon of interest. By focusing on particular statistically relevant features and seeking to explain those features through modeling causal mechanisms, I argue that scientists are able to provide generalized information that then becomes scientific knowledge. According to the Salmon-Roe approach, a combination of statistical relevance and causal process tracing provides scientific explanation; the tracing of intersecting causal processes and the extrapolation of propensities about those causal processes result in scientific explanation. Negative Causation By understanding mechanisms as the practice of tracing causal processes, mystifying cases of negative causation can be modeled and explained. For example, the prevention of dopamine binding to cell receptors, thus causing a raise in cellular prolactin output is something cellular biologists are interested in modeling and explaining. Notice, it is the absence of the binding of dopamine that allows for the production of prolactin by the cell. I argue that we can talk about underlying causal mechanisms in a different way by tracing one or more propagating causal process, possibly through an intersection of those causal processes. If we want to show why Page 1 of 3 Sarah M. Roe sroe@ucdavis.edu Department of Philosophy, University of California, Davis an event occurred, we fill in the relevant processes and interactions that brought us to the event of interest. Revisiting our example, scientists are interested in the intersection of processes that produce an increase in prolactin level within the blood. Process 1 is composed of a particular cell receptor located on a particular lactatrop cell located within the pituitary gland. Process 2 involves a dopamine antagonist (for example, the drug domperidone) in close enough proximity to the particular lactatrop cell. When the processes intersect in spacetime, the drug binds to the particular receptor, successfully inhibiting dopamine from binding to that receptor. So, the phenomenon of interest, for this mechanism is the binding of the drug to the particular receptor, and it is this intersection that results in a continuation of prolactin production by the cell. By modeling mechanisms in this way, we can pick out certain processes that interest us and come to understand how they causally relate to one another. And importantly, the Salmon-Roe approach achieves this without having to reify absences. Negative causation cases are easily remedied when utilizing my notion of mechanisms. Talk of negative causation always signals a different mechanism than we may have first thought. By utilizing the Salmon-Roe account, we can easily identify the relevant interactions and other relevant components in the environment. Once this is understood, we can shift our focus from the seemingly bewildering mechanism and focus on the appropriate mechanism that produces the appropriate phenomenon in question. Scientific Practice Another concern faced by the new mechanists is that their accounts do not mirror scientific practice. More precisely, it is unclear on the new mechanists’ accounts how scientific information transforms from token or particular information into the generalizations required for scientific knowledge. I argue that the Salmon-Roe approach can accurately describe this process. Salmon’s statistical-relevance model of scientific explanation focuses on subsuming facts, or particular information gathered in the world, under generalizations in order to explain. I extend the statistical-relevance model of scientific explanation and focus on how scientists go about arranging particular occurrences of mechanisms and their produced phenomena under general observations and goals of scientific knowledge. I argue that scientists create or construct information taxonomies, or family trees for mechanisms. During this important step, scientists compile their particular findings and also strive to place those findings within a broader scientific context. Once this is accomplished, others can then use those findings either to continue work on the same type of mechanism or for other closely related mechanisms. It is through this iterative process, scientists are able provide useful scientific information, or generalizations, from particular findings. The notion of regularity is yet another concern. Many of the new mechanists insist that mechanisms reliably and regularly produce the phenomenon of interest. That is, mechanisms are supposed to explain why a phenomenon happens most of the time. However, not all mechanisms that scientists are interested in produce regularly occurring phenomena; they might be quite irregular. The Salmon-Roe account is in a better position to reflect this aspect of science because it doesn't require regularity. Some causal mechanisms are just too unreliable and work only once or a few times. In fact, irregular mechanism might be just as important as regular mechanisms. Take, of example, the electrical activity within post-synaptic neurons. Each one fails to release neurotransmitters more often than not. And, moreover, this is both an interesting and important effect. It would be impossible for an organism to function properly if every neuron released neurotransmitters upon receiving an input from another neuron. So, mechanisms that do not reliably produce an intended phenomenon also need explaining and play important roles in nature. As such, the Salmon-Roe account can explain phenomenon that other mechanistic accounts cannot. Mechanistic Environment and Modeling Practices Instead of focusing on entities and activities, I argue that scientists are focused on tracing and modeling processes, precisely because processes have much greater temporal duration and spatial extent than events, and this allows for a better understanding of causality. As a result, my view requires that environment play a more central role in explanation, and thus allows for dynamic modeling practices currently found throughout the biological sciences. On my view, the environment or external factors are included in the causal processes that are being investigated, and the internal/external distinction is pragmatic. So, scientists choose, based on statistically Page 2 of 3 Sarah M. Roe sroe@ucdavis.edu Department of Philosophy, University of California, Davis relevant features for the phenomenon being investigated, what is internal to the mechanism. For example, whether dopamine’s presence in the immediate environment is of importance is determined by scientists, how they construct their models, and the phenomenon being investigated. In this way, some external factors are more prominent because they will actually be considered a part of the mechanism. Rather than thinking of the mechanism as separate from, or different from the environment the mechanism is in, the Salmon-Roe approach makes apparent all features of causal processes and all relevant causal processes. As a result of focusing on processes and promoting the mechanism environment, the Salmon-Roe approach easily allows for dynamic modeling practices. Consider the mechanism that produces the hormone prolactin. Because the mechanism is now considered cyclic, with both negative and positive feedback loops throughout, it is hard to fully understand the mechanism independent of its environment. Notice, the production of prolactin requires entities that are spatiotemporally distant from one another, namely the pituitary gland and lactotropic cells within same organism. The pituitary gland excretes hormones into the blood stream such that, when they come in contact with lactotropic cell receptors, they cause the cell to produce prolactin. The hormone prolactin is then transported out of the cell and into the local blood stream. When the concentration of prolactin within the blood reaches a particular level, as dictated by the pituitary gland, the pituitary gland will then produce a different hormone such that, when it bonds to a lactotropic cell receptor, it halts the production of prolactin by that cell. This cyclic mechanism cannot be thought of as independent of its environment. Moreover, these types of mechanism, ones that interact with the mechanistic environment, are abundant throughout the sciences and also require explanation. By situating dynamic mechanisms within proper environments, I argue that the Salmon-Roe account of mechanistic explanations better reflects model building practices found within contemporary science. Within my dissertation I show how the Salmon-Roe approach to mechanistic explanations can solve problem regarding negative causation, scientific practice, generalities, regularities, complex model building practices, and biological complexity. Throughout, I provide a contemporary case study both to illustrate the depth of these problems for the new mechanists as well as illuminate how the Salmon-Roe account can solve these difficulties. By extending and emphasizing Salmon’s account, the Salmon-Roe account can provide mechanistic explanations throughout the biological sciences by allowing for the modeling of dynamic mechanisms, promoting the role of mechanism environment, providing a way to understand how scientific generalizations are produced, and explaining important types of mechanism like ephemeral mechanisms and cases of negative causation. The result is a novel account of ‘mechanism’ that provides a better understanding of scientific explanation and scientific discovery. Page 3 of 3