It is often claimed—especially by scientific realists—that science provides understanding of the world only if its theories are (at least approximately) true descriptions of reality, in its observable as well as unobservable aspects. This paper critically examines this ‘realist thesis’ concerning understanding. A crucial problem for the realist thesis is that (as study of the history and practice of science reveals) understanding is frequently obtained via theories and models that appear to be highly unrealistic or even completely fictional. So we (...) face the dilemma of either giving up the realist thesis that understanding requires truth, or allowing for the possibility that in many if not all practical cases we do not have scientific understanding. I will argue that the first horn is preferable: the link between understanding and truth can be severed. This becomes a live option if we abandon the traditional view that scientific understanding is a special type of knowledge. While this view implies that understanding must be factive, I avoid this implication by identifying understanding with a skill rather than with knowledge. I will develop the idea that understanding phenomena consists in the ability to use a theory to generate predictions of the target system’s behavior. This implies that the crucial condition for understanding is not truth but intelligibility of the theory, where intelligibility is defined as the value that scientists attribute to the theoretical virtues that facilitate the construction of models of the phenomena. I will show, first, that my account accords with the way practicing scientists conceive of understanding, and second, that it allows for the use of idealized or fictional models and theories in achieving understanding. (shrink)

In a recent paper, Kaplan (Synthese 183:339–373, 2011) takes up the task of extending Craver’s (Explaining the brain, 2007) mechanistic account of explanation in neuroscience to the new territory of computational neuroscience. He presents the model to mechanism mapping (3M) criterion as a condition for a model’s explanatory adequacy. This mechanistic approach is intended to replace earlier accounts which posited a level of computational analysis conceived as distinct and autonomous from underlying mechanistic details. In this paper I discuss work in (...) computational neuroscience that creates difficulties for the mechanist project. Carandini and Heeger (Nat Rev Neurosci 13:51–62, 2012) propose that many neural response properties can be understood in terms of canonical neural computations. These are “standard computational modules that apply the same fundamental operations in a variety of contexts.” Importantly, these computations can have numerous biophysical realisations, and so straightforward examination of the mechanisms underlying these computations carries little explanatory weight. Through a comparison between this modelling approach and minimal models in other branches of science, I argue that computational neuroscience frequently employs a distinct explanatory style, namely, efficient coding explanation. Such explanations cannot be assimilated into the mechanistic framework but do bear interesting similarities with evolutionary and optimality explanations elsewhere in biology. (shrink)

This paper addresses concerns raised recently by Datteri (Biol Philos 24:301–324, 2009) and Craver (Philos Sci 77(5):840–851, 2010) about the use of brain-extending prosthetics in experimental neuroscience. Since the operation of the implant induces plastic changes in neural circuits, it is reasonable to worry that operational knowledge of the hybrid system will not be an accurate basis for generalisation when modelling the unextended brain. I argue, however, that Datteri’s no-plasticity constraint unwittingly rules out numerous experimental paradigms in behavioural and systems (...) neuroscience which also elicit neural plasticity. Furthermore, I propose that Datteri and Craver’s arguments concerning the limitations of prosthetic modelling in basic neuroscience, as opposed to neuroengineering, rests on too narrow a view of the ways models in neuroscience should be evaluated, and that a more pluralist approach is needed. I distinguish organisational validity of models from mechanistic validity. I argue that while prosthetic models may be deficient in the latter of these explanatory virtues because of neuroplasticity, they excel in the former since organisational validity tracks the extent to which a model captures coding principles that are invariant with plasticity. Changing the brain, I conclude, is one viable route towards explaining the brain. (shrink)

The concept of the receptive field, first articulated by Hartline, is central to visual neuroscience. The receptive field of a neuron encompasses the spatial and temporal properties of stimuli that activate the neuron, and, as Hubel and Wiesel conceived of it, a neuron’s receptive field is static. This makes it possible to build models of neural circuits and to build up more complex receptive fields out of simpler ones. Recent work in visual neurophysiology is providing evidence that the classical receptive (...) field is an inaccurate picture. The receptive field seems to be a dynamic feature of the neuron. In particular, the receptive field of neurons in V1 seems to be dependent on the properties of the stimulus. In this paper, we review the history of the concept of the receptive field and the problematic data. We then consider a number of possible theoretical responses to these data. (shrink)

Woodward's long awaited book is an attempt to construct a comprehensive account of causation explanation that applies to a wide variety of causal and explanatory claims in different areas of science and everyday life. The book engages some of the relevant literature from other disciplines, as Woodward weaves together examples, counterexamples, criticisms, defenses, objections, and replies into a convincing defense of the core of his theory, which is that we can analyze causation by appeal to the notion of manipulation.

In this sequence of philosophical essays about natural science, the author argues that fundamental explanatory laws, the deepest and most admired successes of modern physics, do not in fact describe regularities that exist in nature. Cartwright draws from many real-life examples to propound a novel distinction: that theoretical entities, and the complex and localized laws that describe them, can be interpreted realistically, but the simple unifying laws of basic theory cannot.

Truth is standardly considered a requirement on epistemic acceptability. But science and philosophy deploy models, idealizations and thought experiments that prescind from truth to achieve other cognitive ends. I argue that such felicitous falsehoods function as cognitively useful fictions. They are cognitively useful because they exemplify and afford epistemic access to features they share with the relevant facts. They are falsehoods in that they diverge from the facts. Nonetheless, they are true enough to serve their epistemic purposes. Theories that contain (...) them have testable consequences, hence are factually defeasible. (shrink)

Empirical adequacy, formal explanation and understanding are distinct goals of science. While no a priori criterion for understanding should be laid down, there may be inherent limitations on the way we are able to understand explanations of physical phenomena. I examine several recent contributions to the exercise of fashioning an explanatory discourse to mold the formal explanation provided by quantum mechanics to our modes of understanding. The question is whether we are capable of truly understanding (or comprehending) quantum phenomena, as (...) opposed to simply accepting the formalism and certain irreducible quantum correlations. The central issue is that of understanding versus merely redefining terms to paper over our ignorance. (shrink)

Scientists depend on complex computational systems that are often ineliminably opaque, to the detriment of our ability to give scientific explanations and detect artifacts. Some philosophers have s...

In artificial intelligence, recent research has demonstrated the remarkable potential of Deep Convolutional Neural Networks (DCNNs), which seem to exceed state-of-the-art performance in new domains weekly, especially on the sorts of very difficult perceptual discrimination tasks that skeptics thought would remain beyond the reach of artificial intelligence. However, it has proven difficult to explain why DCNNs perform so well. In philosophy of mind, empiricists have long suggested that complex cognition is based on information derived from sensory experience, often appealing to (...) a faculty of abstraction. Rationalists have frequently complained, however, that empiricists never adequately explained how this faculty of abstraction actually works. In this paper, I tie these two questions together, to the mutual benefit of both disciplines. I argue that the architectural features that distinguish DCNNs from earlier neural networks allow them to implement a form of hierarchical processing that I call “transformational abstraction”. Transformational abstraction iteratively converts sensory-based representations of category exemplars into new formats that are increasingly tolerant to “nuisance variation” in input. Reflecting upon the way that DCNNs leverage a combination of linear and non-linear processing to efficiently accomplish this feat allows us to understand how the brain is capable of bi-directional travel between exemplars and abstractions, addressing longstanding problems in empiricist philosophy of mind. I end by considering the prospects for future research on DCNNs, arguing that rather than simply implementing 80s connectionism with more brute-force computation, transformational abstraction counts as a qualitatively distinct form of processing ripe with philosophical and psychological significance, because it is significantly better suited to depict the generic mechanism responsible for this important kind of psychological processing in the brain. (shrink)

Science relies on models and idealizations that are known not to be true. Even so, science is epistemically reputable. To accommodate science, epistemology should focus on understanding rather than knowledge and should recognize that the understanding of a topic need not be factive. This requires reconfiguring the norms of epistemic acceptability. If epistemology has the resources to accommodate science, it will also have the resources to show that art too advances understanding.

Science is the study of our world, as it is in its messy reality. Nonetheless, science requires idealization to function—if we are to attempt to understand the world, we have to find ways to reduce its complexity. Idealization and the Aims of Science shows just how crucial idealization is to science and why it matters. Beginning with the acknowledgment of our status as limited human agents trying to make sense of an exceedingly complex world, Angela Potochnik moves on to explain (...) how science aims to depict and make use of causal patterns—a project that makes essential use of idealization. She offers case studies from a number of branches of science to demonstrate the ubiquity of idealization, shows how causal patterns are used to develop scientific explanations, and describes how the necessarily imperfect connection between science and truth leads to researchers’ values influencing their findings. The resulting book is a tour de force, a synthesis of the study of idealization that also offers countless new insights and avenues for future exploration. (shrink)

Many of the computing systems programmed using Machine Learning are opaque: it is difficult to know why they do what they do or how they work. Explainable Artificial Intelligence aims to develop analytic techniques that render opaque computing systems transparent, but lacks a normative framework with which to evaluate these techniques’ explanatory successes. The aim of the present discussion is to develop such a framework, paying particular attention to different stakeholders’ distinct explanatory requirements. Building on an analysis of “opacity” from (...) philosophy of science, this framework is modeled after accounts of explanation in cognitive science. The framework distinguishes between the explanation-seeking questions that are likely to be asked by different stakeholders, and specifies the general ways in which these questions should be answered so as to allow these stakeholders to perform their roles in the Machine Learning ecosystem. By applying the normative framework to recently developed techniques such as input heatmapping, feature-detector visualization, and diagnostic classification, it is possible to determine whether and to what extent techniques from Explainable Artificial Intelligence can be used to render opaque computing systems transparent and, thus, whether they can be used to solve the Black Box Problem. (shrink)

From antiquity to the end of the twentieth century, philosophical discussions of understanding remained undeveloped, guided by a 'received view' that takes understanding to be nothing more than knowledge of an explanation. More recently, however, this received view has been criticized, and bold new philosophical proposals about understanding have emerged in its place. In this book, Kareem Khalifa argues that the received view should be revised but not abandoned. In doing so, he clarifies and answers the most central questions in (...) this burgeoning field of philosophical research: what kinds of cognitive abilities are involved in understanding? What is the relationship between the understanding that explanations provide and the understanding that experts have of broader subject matters? Can there be understanding without explanation? How can one understand something on the basis of falsehoods? Is understanding a species of knowledge? What is the value of understanding? (shrink)

Bionic systems, connecting biological tissues with computer or robotic devices through brain–machine interfaces, can be used in various ways to discover biological mechanisms. In this article I outline and discuss a “stimulation-connection” bionics-supported methodology for the study of the brain, and compare it with other epistemic uses of bionic systems described in the literature. This methododology differs from the “synthetic”, simulative method often followed in theoretically driven Artificial Intelligence and cognitive science, even though it involves machine models of biological systems. (...) I also bring the previous analysis to bear on some claims on the epistemic value of bionic technologies made in the recent philosophical literature. I believe that the methodological reflections proposed here may contribute to the piecewise understanding of the many ways bionic technologies can be deployed not only to restore lost sensory-motor functions, but also to discover brain mechanisms. (shrink)

Understanding is a central aim of science and highly important in present-day society. But what precisely is scientific understanding and how can it be achieved? This book answers these questions, through philosophical analysis and historical case studies, and presents a philosophical theory of scientific understanding that highlights its contextual nature.

What are the relative epistemic merits of building prosthetic models versus building nonprosthetic models and simulations? I argue that prosthetic models provide a sufficient test of affordance validity, that is, of whether the target system affords mechanisms that can be commandeered by a prosthesis. In other respects, prosthetic models are epistemically on par with nonprosthetic models. I focus on prosthetics in neuroscience, but the results are general. The goal of understanding how brain mechanisms work under ecologically and physiologically relevant conditions (...) is narrow compared to the search for maker's knowledge about how the brain can be made to work for us. (shrink)

There is a vast literature within philosophy of mind that focuses on artificial intelligence, but hardly mentions methodological questions. There is also a growing body of work in philosophy of science about modeling methodology that hardly mentions examples from cognitive science. Here these discussions are connected. Insights developed in the philosophy of science literature about the importance of idealization provide a way of understanding the neural implausibility of connectionist networks. Insights from neurocognitive science illuminate how relevant similarities between models and (...) targets are picked out, how modeling inferences are justified, and the metaphysical status of models. (shrink)

We argue that dynamical and mathematical models in systems and cognitive neuro- science explain (rather than redescribe) a phenomenon only if there is a plausible mapping between elements in the model and elements in the mechanism for the phe- nomenon. We demonstrate how this model-to-mechanism-mapping constraint, when satisfied, endows a model with explanatory force with respect to the phenomenon to be explained. Several paradigmatic models including the Haken-Kelso-Bunz model of bimanual coordination and the difference-of-Gaussians model of visual receptive fields are (...) explored. (shrink)

With In Search of Mechanisms, Carl F. Craver and Lindley Darden offer both a descriptive and an instructional account of how biologists discover mechanisms. Drawing on examples from across the life sciences and through the centuries, Craver and Darden compile an impressive toolbox of strategies that biologists have used and will use again to reveal the mechanisms that produce, underlie, or maintain the phenomena characteristic of living things. They discuss the questions that figure in the search for mechanisms, characterizing the (...) experimental, observational, and conceptual considerations used to answer them, all the while providing examples from the history of biology to highlight the kinds of evidence and reasoning strategies employed to assess mechanisms. At a deeper level, Craver and Darden pose a systematic view of what biology is, of how biology makes progress, of how biological discoveries are and might be made, and of why knowledge of biological mechanisms is important for the future of the human species. (shrink)

This article examines three candidate cases of non-causal explanation in computational neuroscience. I argue that there are instances of efficient coding explanation that are strongly analogous to examples of non-causal explanation in physics and biology, as presented by Batterman, Woodward, and Lange. By integrating Lange’s and Woodward’s accounts, I offer a new way to elucidate the distinction between causal and non-causal explanation, and to address concerns about the explanatory sufficiency of non-mechanistic models in neuroscience. I also use this framework to (...) shed light on the dispute over the interpretation of dynamical models of the brain. _1_ Introduction _1.1_ Efficient coding explanation in computational neuroscience _1.2_ Defining non-causal explanation _2_ Case I: Hybrid Computation _3_ Case II: The Gabor Model Revisited _4_ Case III: A Dynamical Model of Prefrontal Cortex _4.1_ A new explanation of context-dependent computation _4.2_ Causal or non-causal? _5_ Causal and Non-causal: Does the Difference Matter? (shrink)

Bionic technologies connecting biological nervous systems to computer or robotic devices for therapeutic purposes have been recently claimed to provide novel experimental tools for the investigation of biological mechanisms. This claim is examined here by means of a methodological analysis of bionics-supported experimental inquiries on adaptive sensory-motor behaviours. Two broad classes of bionic systems (regarded here as hybrid simulations of the target biological system) are identified, which differ from each other according to whether a component of the biological target system (...) is replaced by an artificial component, or else a component of an artificial system is replaced by a biological component. The role of these hybrid systems in the modelling of adaptive sensory-motor biological behaviours is discussed with reference to bionics-supported experiments on the mechanisms of body stabilization in lampreys. Methodological problems emerging from these case studies often arise in computer-based and biorobotic simulations of biological behaviours too. Accordingly, the present analysis contributes to identifying a more general regulative methodological framework for the machine-based modelling of biological systems. (shrink)

There has been considerable debate in the literature about the relative merits of information processing versus dynamical approaches to understanding cognitive processes. In this article, we explore the relationship between these two styles of explanation using a model agent evolved to solve a relational categorization task. Specifically, we separately analyze the operation of this agent using the mathematical tools of information theory and dynamical systems theory. Information-theoretic analysis reveals how task-relevant information flows through the system to be combined into a (...) categorization decision. Dynamical analysis reveals the key geometrical and temporal interrelationships underlying the categorization decision. Finally, we propose a framework for directly relating these two different styles of explanation and discuss the possible implications of our analysis for some of the ongoing debates in cognitive science. (shrink)