Idealizations in Science: Exploration, Understanding, and Confirmation.
Idealizations are ubiquitous in science. But in appealing to idealizations, abstractions, approximations, and metaphors, scientists are making use of falsehoods and distortions, and they are doing so in order to tell us truths about the world. Some of the most extreme of such falsehoods are “infinite idealizations,” such as models with infinite populations in biology or infinitely divisible goods in economics. Moreover, indispensable infinite idealizations can lead to paradoxes. For instance, while it is known that boiling kettles are finite in size, our best theories of boiling—a king of thermodynamic phase transition—describe kettles as infinite systems. Questions arise: How and why do scientists appeal to lies (in the way of idealizations and approximations) in order to search for and discover truths? What are we to make of ostensibly indispensable and infinite idealizations, and the paradoxes associated with them? What roles do they play and how are they justified? And what implications do important case studies involving infinite idealizations have for the main issues in philosophy of science, viz., scientific explanation and understanding, reduction and emergence, scientific realism and mathematical Platonism, and confirmation and evidence?
I am working on answering such questions in a series of paper and in two book-length projects. First, my book Idealizations in Physics (CUP) gives an up-to-date review of the literature and sets the agenda for future work. I argue that there is a need to attend to an epistemic-logical justificatory problem associated with idealizations, the role played by in-principle de-idealization and the in-principle/in-practice distinction in solving said problem, and the difficulties associated with providing the relevant justification and solution. Difficulties arise partly in light of the fact that the nature of idealization is entangled with one’s commitments to epistemic and ontological views of science and the world, including issues such as interpretation of background theories, modality, scale and levels reality, as well as induction and confirmation. If the justificatory problem can be solved while maintaining the indispensability of idealizations, and a paradox in the way of imputing contradictory properties to target systems can be avoided, then I have suggested that realists may need to commit to stronger views than they intended, or else embrace acceptable anti-realist positions. Additionally, the book sketches the idea that that idealizations play a role in understanding-what some phenomena are in the first place and in understanding the structure of theory in light of said phenomena.
The second book-length project—tentatively titled Infinite Idealizations in Science: Exploration, Understanding, and Confirmation—is a continuation of the first and concerns a philosophical analysis of the structure and implications of infinite idealizations in science (including case studies from physics, biology, economics and statistics). My aim is to critically analyze the debate in the scientific and philosophical literature revolving around infinite and infinitesimal idealizations in science and argue for (at least) three novel theses: First, that infinite idealizations play an indispensable exploratory (as opposed to explanatory) role in science, which is necessary for a deep sense of (modal) understanding of scientific theories and models, intertheoretic relations, and physical phenomena. Second, that paradoxical infinite/infinitesimal idealizations sustain a notion of strong emergence that signals the need for novel physical-mathematical research programs. Last, that infinite idealizations sometimes play an essential role in cross-model justification and confirmation, and in cross-disciplinary knowledge transfer. In relation to the last thesis, by looking specifically to examples from condensed matter physics that involve multiscale analysis, I have argued for a scale-dependent conception of ontology and I plan to further develop this line of thought. (Part of this work is based on collaboration with Axel Gelfert and Patrick McGivern.)
Philosophy of Biology and Mitonuclear Ecology: Evolutionary Fitness and Species Concepts
In a collaborative project with biologist Kyle Heine, we are developing a mitonuclear approach to philosophy of biology in which we apply insights from the emerging field of mitonuclear ecology (having to do with interactions of mitochondrial and nuclear genomes) to concepts such as fitness, adaption, speciation, etc. For example, in paper published in Biology and Philosophy, we look to mitonuclear ecology and the phenomenon of Mother’s Curse to argue that the sex of parents and offspring among populations of eukaryotic organisms, as well as the mitochondrial genome, ought to be taken into account in the conceptualization of evolutionary fitness. We show how characterizations of fitness considered by philosophers that do not take sex and the mitochondrial genome into account suffer and, consequently, we suggest that the concept of trait fitness is more fundamental than that of organism fitness.
Philosophy of Machine Learning: Artificial Understanding, Modeling, and Inference
Advances in Artificial Intelligence and Machine Learning (ML), especially using Deep Learning (DL) techniques, have accelerated performance in numerous areas of practical application. In numerous instances DL has enabled algorithms to approach and overtake human performance such as AlphaGo’s defeat of Lee Sedol in 2016. Interestingly, human competitive performance on such benchmarks has accompanied an increased use of terms like “understanding” in artificial contexts. In a collaborative project with data scientist Michael Tamir, we aim to flesh out some of the philosophical consequences of such recent developments. For example, in a current project, we identify practical ability and reliable task performance, as well as information compression and representation, as key factors indicative of understanding in the philosophical literature. We then explore these factors in the context of DL algorithms, recognizing prominent patterns in how the algorithms represent information. While DL applications do not qualify as the kind of mental activity found in persons, we argue that coupling analyses from philosophical accounts with the empirical and theoretical basis for identifying these factors in DL representations provides a framework for discussing potential artificial understanding.
In addition, reflecting on how ML trained algorithms are often called “models,” in two forthcoming papers we explain how such automated effective estimation algorithms fit in with existing accounts of how scientists leverage modeling and representation for understanding. We suggest a candidate for framing the proper target of ML understanding and identify three modes of understanding given the said proposed target of ML models. In a further current project, making a connection with the philosophy of statistics literature we attempt to extend Deborah Mayo and Aris Spanos’ Error-Statistical Approach (ESA) to statistical modeling and their emphasis on reliable inductive inference (RII) by proposing that that ML data validation and testing potentially provides a direct, measurable method of meeting RII in contexts where some ESA assumptions cannot be met.
Metaphysics of Color and Objectivism
In a collaborative book—tentatively titled Metaphysics of Color (under contract with CUP)—with metaphysician Michael Watkins, we are working on defending an objectivist theory of colors, in which colors are objective properties of objects. Such a position is in contrast to current well-received views such as that of relationalism or eliminativism about colors. In a recent publication we first note how arguments from perceptual variation challenge the view that colors are objective properties of objects, and we diagnose one central reason why arguments from perceptual variation seem especially challenging for objectivists about color. Next, we offer a response to this challenge, claiming that once we focus on determinate colors rather than the determinables they determine, a reply to arguments from perceptual variation becomes apparent. Third, our nominal opponents are relationalists about color and we will argue that the main argument for rejecting objectivism commits the relationalist to a position that is more radical than the one she would wish to endorse. Fourth, we suggest that insight into which properties could be relational may be found by looking to our best scientific theories.
In the current book project, we place our arguments in the context of the larger literature on the metaphysics of color, reflecting on theories of color such eliminativism, relationalism, and orthodox objectivism, and present an alternative objectivist account of colors. Distinguishing between (what we call) the color problem and the color perception problem, we argue that we can solve the former without solving the latter, viz., that we determine visually what color an object has without determining which color experiences are veridical. We also present (what we call) the problem of perceptual agreement (of determinate color) and submit that it proves to be a serious and long overlooked obstacle for those insisting that colors are not objective features of objects, viz., non-objectivist theories like eliminativism and relationalism. Last, we explore some connections between objectivism about color and recent work in the philosophy of science by thinking about colors as natural kind properties, scale relative real patterns, and explanatorily indispensable components of our best scientific theories.
Epistemology of Science: The Material Theory of Induction & Consequences
John D. Norton has recently defended a material theory of induction (MTI) that describe the evidential relations between the propositions constituting the content of science. Part of my work develops applications of the MTOI to debates in philosophy concerning metaphysical issues including scientific realism/anti-realism, physicalism, and mathematical platonism, as well as issues in epistemology including internalist vs. externalist concepts of justification, and learning from data. The basic idea is to try to flesh out which ontological commitments we ought to have if we think that science works and gains knowledge via induction as specified by the MTOI, and to further understand the type of epistemology supported by the MTI. For instance, I have argued that arguments based on historical inductions such as the pessimistic meta-induction and the problem of unconceived alternatives ought to be rejected as non-cogent arguments when analyzed via the MTOI. There are clear consequences for other areas that make use of historical inductions, e.g., standard positive arguments for physicalism appeal to such inductions. I have also suggested that the MTI can support a local (instead of global/wholesale) realism/empiricism about particular unobservable entities and behaviors
Science and Art: Representation, Fictions, and the Imagination
There are interesting analogies to be drawn between the literature in philosophy of science and the literature in aesthetics and philosophy of art on issues such as representation, fictions, and the imagination. Part of my research consists in identifying such analogies, as well as making connections with thought experiments and impossible objects.
On the Back Burner
Coulomb’s Torsion Balance, Galileo’s Inclined Plane, and the Philosophy of Experimentation
I have recreated various important experiments in the history of science such as Galileo’s inclined plane along with colleagues Eric Hatleback and Paolo Palmieri in order to gain insight into the history and philosophy of experimentation. The main project that I spearheaded includes inquiring into the material intricacies of Charles Augustin Coulomb’s electric torsion balance experiment, which led to the famous “Coulomb’s Law” in electrostatics. Specifically, contemporary scholars are engaged in a debate over whether the results that Coulomb presented in his 1785 and 1787 memoirs to the Paris Academy of Sciences were attained experimentally or theoretically. We study Coulomb’s famous 1785 electric torsion balance experiment through analysis of relevant texts and, more importantly, through a replication that is more faithful to Coulomb’s original design than previous attempts. We show that, despite recent claims, (1) it has so far proved impossible to obtain the same results reported by Coulomb in his paper of 1785, (2) Coulomb’s published results are most likely atypical, and (3) electric torsion balance experiments degenerate quickly when parameters are altered by small amounts. Preprint: http://philsci-archive.pitt.edu/11048