The medium‐independence of computational descriptions has shaped common conceptions of computational explanation. So long as our goal is to explain how a system successfully carries out its computations, then we only need to describe the abstract series of operations that achieve the desired input–output mapping, however they may be implemented. It is argued that this abstract conception of computational explanation cannot be applied to so‐called real‐time computing systems, in which meeting temporal deadlines imposed by the systems with which a device (...) interfaces are constitutive of the computing tasks that a device performs. Instead, real‐time computing reveals the need for alternative conceptions of computational explanation, as well as computational implementation, that eschew medium‐independence. (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)

Computational explanations focus on information processing required in specific cognitive capacities, such as perception, reasoning or decision-making. These explanations specify the nature of the information processing task, what information needs to be represented, and why it should be operated on in a particular manner. In this article, the focus is on three questions concerning the nature of computational explanations: What type of explanations they are, in what sense computational explanations are explanatory and to what extent they involve a special, “independent” (...) or “autonomous” level of explanation. In this paper, we defend the view computational explanations are genuine explanations, which track non-causal/formal dependencies. Specifically, we argue that they do not provide mere sketches for explanation, in contrast to what for example Piccinini and Craver :283–311, 2011) suggest. This view of computational explanations implies some degree of “autonomy” for the computational level. However, as we will demonstrate that does not make this view “computationally chauvinistic” in a way that Piccinini or Kaplan :339–373, 2011) have charged it to be. (shrink)

This paper deals with the question: what are the key requirements for a physical system to perform digital computation? Time and again cognitive scientists are quick to employ the notion of computation simpliciter when asserting basically that cognitive activities are computational. They employ this notion as if there was or is a consensus on just what it takes for a physical system to perform computation, and in particular digital computation. Some cognitive scientists in referring to digital computation simply adhere to (...) Turing’s notion of computability . Classical computability theory studies what functions on the natural numbers are computable and what mathematical problems are undecidable. Whilst a mathematical formalism of computability may perform a methodological function of evaluating computational theories of certain cognitive capacities, concrete computation in physical systems seems to be required for explaining cognition as an embodied phenomenon . There are many non-equivalent accounts of digital computation in physical systems. I examine only a handful of those in this paper: (1) Turing’s account ; (2) The triviality “account”; (3) Reconstructing Smith’s account of participatory computation ; (4) The Algorithm Execution account . My goal in this paper is twofold. First, it is to identify and clarify some of the underlying key requirements mandated by these accounts. I argue that these differing requirements justify a demand that one commits to a particular account when employing the notion of computation in regard to physical systems. Second, it is to argue that despite the informative role that mathematical formalisms of computability may play in cognitive science, they do not specify the relationship between abstract and concrete computation. (shrink)

The duality of truth and falsity in a Boolean algebra of propositions is used to generate a duality of belief and disbelief. To each additive probability measure that represents belief there corresponds a dual additive measure that represents disbelief. The dual measure has its own peculiar calculus, in which, for example, measures are added when propositions are combined under conjunction. A Venn diagram of the measure has the contradiction as its total space. While additive measures are not self-dual, the epistemic (...) state of complete ignorance is represented by the unique, monotonic, non-additive measure that is self-dual in its contingent propositions. Convex sets of additive measures fail to represent complete ignorance since they are not self-dual. (shrink)

In Origins of Objectivity Burge advances a theory of perception according to which perceptions are, themselves, objective representations. The possession of veridicality conditions by perceptual states—roughly, non-propositional analogues of truth-conditions—is central to Burge’s account of how perceptual states differ, empirically and metaphysically, from sensory states. Despite an impressive examination of the relevant empirical literatures, I argue here that Burge has not succeeded in securing a distinction between perception and “mere” sensation.

The Doomsday argument and anthropic reasoning are two puzzling examples of probabilistic confirmation. In both cases, a lack of knowledge apparently yields surprising conclusions. Since they are formulated within a Bayesian framework, they constitute a challenge to Bayesianism. Several attempts, some successful, have been made to avoid these conclusions, but some versions of these arguments cannot be dissolved within the framework of orthodox Bayesianism. I show that adopting an imprecise framework of probabilistic reasoning allows for a more adequate representation of (...) ignorance in Bayesian reasoning and explains away these puzzles. (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)

The purpose of this paper is to defend, contra Fodor and Piattelli-Palmarini (F&PP), that the theory of natural selection (NS) is a perfectly bona fide empirical unified explanatory theory. F&PP claim there is nothing non-truistic, counterfactual-supporting, of an “adaptive” character and common to different explanations of trait evolution. In his debate with Fodor, and in other works, Sober defends NS but claims that, compared with classical mechanics (CM) and other standard theories, NS is peculiar in that its explanatory models are (...) a priori (a trait shared with few other theories). We argue that NS provides perfectly bona fide adaptive explanations of phenotype evolution, unified by a common natural-selection guiding principle. First, we introduce the debate and reply to F&PP’s main argument against NS. Then, by reviewing different examples and analyzing Fisher’s model in detail, we show that NS explanations of phenotypic evolution share a General Natural Selection Principle. Third, by elaborating an analogy with CM, we argue against F&PP’s claim that such a principle would be a mere truism and thus explanatorily useless, and against Sober’s thesis that NS models/explanations have a priori components that are not present in CM and other common empirical theories. Irrespective of differences in other respects, the NS guiding principle has the same epistemic status as other guiding principles in other highly unified theories such as CM. We argue that only by pointing to the guiding principle-driven nature that it shares with CM and other highly unified theories, something no-one has done yet in this debate, one can definitively show that NS is not defective in F&PP’s sense: in the respects relevant to the debate, Natural Selection is as defective and as epistemically peculiar as Classical Mechanics and other never questioned theories. (shrink)

Descriptive abstraction means omission of information from descriptions of phenomena. In this paper, I introduce a distinction between vertical and horizontal descriptive abstraction. Vertical abstracts away levels of mechanism or organization, while horizontal abstracts away details within one level of organization. The distinction is implicit in parts of the literature, but it has received insufficient attention and gone mainly unnoticed. I suggest that the distinction can be used to clarify how computational descriptions are formed in some variants of the mechanistic (...) account of physical computation. Furthermore, I suggest that, if this suggestion is adopted, it can be used to resolve what I call abstraction, hierarchy, and generality problems raised against mechanistic account of physical computation. According to the abstraction problem, the mechanistic account of physical computation is conceptually confused in claiming that physical systems process computational, abstract properties. An existing solution distinguishes between descriptive and metaphysical abstraction, suggesting that the abstraction problem unnecessarily postulates metaphysically abstract entities. The solution has been criticized for leading to what I call hierarchy and generality problems: it results in two separate hierarchies, one physical and one computational, making it problematic both to account for the generality of computational descriptions and to specify how the two hierarchies are related to each other. Adopting the vertical-horizontal distinction and the view that computational descriptions are achieved by horizontal abstraction allows one to account for the generality of computational descriptions, and to form a single hierarchy in which there are no separate hierarchies in need of integration. (shrink)

The mechanistic view of computation contends that computational explanations are mechanistic explanations. Mechanists, however, disagree about the precise role that the environment – or the so-called “contextual level” – plays for computational explanations. We advance here two claims: Contextual factors essentially determine the computational identity of a computing system ; this means that specifying the “intrinsic” mechanism is not sufficient to fix the computational identity of the system. It is not necessary to specify the causal-mechanistic interaction between the system and (...) its context in order to offer a complete and adequate computational explanation. While the first claim has been discussed before, the second has been practically ignored. After supporting these claims, we discuss the implications of our contextualist view for the mechanistic view of computational explanation. Our aim is to show that some versions of the mechanistic view are consistent with the contextualist view, whilst others are not. (shrink)

Computational neuroscientists not only employ computer models and simulations in studying brain functions. They also view the modeled nervous system itself as computing. What does it mean to say that the brain computes? And what is the utility of the ‘brain-as-computer’ assumption in studying brain functions? In previous work, I have argued that a structural conception of computation is not adequate to address these questions. Here I outline an alternative conception of computation, which I call the analog-model. The term ‘analog-model’ (...) does not mean continuous, non-discrete or non-digital. It means that the functional performance of the system simulates mathematical relations in some other system, between what is being represented. The brain-as-computer view is invoked to demonstrate that the internal cellular activity is appropriate for the pertinent information-processing task.Keywords: Computation; Computational neuroscience; Analog computers; Representation; Simulation. (shrink)

What does it mean to say that a physical system computes or, specifically, to say that the nervous system computes? One answer, endorsed here, is that computing is a sort of modeling. I trace this line of answer in the conceptual and philosophical work conducted over the last 3 decades by researchers associated with the University of California, San Diego. The linkage between their work and the modeling notion is no coincidence: the modeling notion aims to account for the computational (...) approach in neuroscience, and UCSD has been home to central studies in neurophilosophy, connectionism, and computational neuroscience. (shrink)

ABSTRACTThe seminal work of David Marr, popularized in his classic work Vision, continues to exert a major influence on both cognitive science and philosophy. The interpretation of his work also co...

Are all three of Marr's levels needed? Should they be kept distinct? We argue for the distinct contributions and methodologies of each level of analysis. It is important to maintain them because they provide three different perspectives required to understand mechanisms, especially information-processing mechanisms. The computational perspective provides an understanding of how a mechanism functions in broader environments that determines the computations it needs to perform. The representation and algorithmic perspective offers an understanding of how information about the environment is (...) encoded within the mechanism and what are the patterns of organization that enable the parts of the mechanism to produce the phenomenon. The implementation perspective yields an understanding of the neural details of the mechanism and how they constrain function and algorithms. Once we adequately characterize the distinct role of each level of analysis, it is fairly straightforward to see how they relate. (shrink)

A theoretical pillars of vision science in the information-processing tradition is that perception involves unconscious inference. The classic support for this claim is that, since retinal inputs underdetermine their distal causes, visual perception must be the conclusion of a process that starts with premises representing both the sensory input and previous knowledge about the visible world. Focus on this “argument from underdetermination” gives the impression that, if it fails, there is little reason to think that visual processing involves unconscious inference. (...) Here an alternative means of support for this pillar is proposed, based on another foundational challenge for the visual system: recognizing invariant properties of objects in the environment even though anything we encounter is never seen exactly the same way twice. Explaining how the visual system solves this invariance problem requires positing visual processes that exhibit many commonalities with inductive inference. Thus, this novel “argument from invariance” reveals one way in which visual processing clearly involves unconscious inference. (shrink)

The Gricean paradigm in pragmatics has recently been attacked for its alleged lack of explanatory import, based on the claim that it does not seek accounts of how utterance interpretation actually works, but merely of how it might work. This article rebuts this line of attack by offering a clear and detailed account of the explanatory project of Gricean pragmatics according to which the latter aims for rationalizing explanations of utterance interpretation. It is shown that, on this view, Gricean pragmatics (...) seeks psychological explanations of utterance interpretation that are “cognitively real” in a perfectly clear and robust sense. (shrink)

Theories in cognitive science, and especially cognitive neuroscience, often claim that parts of cognitive systems are reused for different cognitive functions. Philosophical analysis of this concept, however, is rare. Here, I first provide a set of criteria for an analysis of reuse, and then I analyse reuse in terms of the functions of subsystems. I also discuss how cognitive systems execute cognitive functions, the relation between learning and reuse, and how to differentiate reuse from related concepts like multi-use, redundancy, and (...) duplication of parts. Finally, I illustrate how my account captures the reuse of dynamical subsystems as unveiled by recent research in cognitive neurobiology. This recent research suggests two different evolutionary justifications for reuse claims. _1_ Introduction _2_ Criteria for Analyses of Reuse _3_ Use and Reuse _4_ The Reuse of Dynamical Systems in Cognitive Systems _5_ Conclusion. (shrink)

The aim of this paper is to present the ontic approach to the normativity of cognitive functions and mechanisms, which is directly related to the understanding of biological normativity in terms of normative mechanisms. This approach assumes the hypothesis that cognitive processes contain a certain normative component independent of external attributions and researchers’ beliefs. This component consists of specific cognitive mechanisms, which I call normative. I argue that a mechanism is normative when it constitutes given actions or behaviors of a (...) system. More precisely, it means that, on the one hand, it is their constitutive cause, and on the other hand, it determines a certain field of possibilities from which the system, guided by its own goals, preferences, environmental constraints, etc., chooses the appropriate action or behavior according to a given situation. The background for the analyses presented here is the predictive processing framework, in which it can be shown that at least some of the predictive mechanisms are in fact normative mechanisms. I refer here to the existence of a motivational relation which determines the normative dependence of the agent’s actions due to specific predictions and environmental constraints. (shrink)

Building on an overview of dual systems theories in behavioural economics, the paper presents a methodological assessment in terms of the mechanistic explanations framework that has gained prominence in philosophy of the neurosciences. I conclude that they fail to meet the standards of causal explanations and I suggest an alternative ‘dual functions’ view based on Marr’s methodology of computational neuroscience. Recent psychological and neuroscience research undermines the case for a categorization of brain processes in terms of properties such as relative (...) speed. I defend an interpretation of dualities as functional, without assigning them to specific neurophysiological structures. (shrink)

While the research programs in early cognitive science and artificial intelligence aimed to articulate what cognition was in ideal terms, much research in contemporary computational neuroscience looks at how and why brains fail to function as they should ideally. This focus on impairment affects how we understand David Marr's hypothesized three levels of understanding. In this essay, we suggest some refinements to Marr's distinctions using a population activity model of cortico-striatal circuitry exploring impulsivity and behavioral inhibition as a case study. (...) In particular, we urge that Marr's computational level should be redefined to include a description of how systems break down. We also underscore that feed-forward processing, cognition disconnected from behavioral context, and representations do not always drive cognition in the way that Marr originally assumed. (shrink)

According to the semantic view of computation, computations cannot be individuated without invoking semantic properties. A traditional argument for the semantic view is what we shall refer to as the argument from the cognitive science practice. In its general form, this argument rests on the idea that, since cognitive scientists describe computations (in explanations and theories) in semantic terms, computations are individuated semantically. Although commonly invoked in the computational literature, the argument from the cognitive science practice has never been discussed (...) in detail. In this paper, we shall provide a critical reconstruction of this argument and an extensive analysis of its prospects, taking into account some ways of defending it that have never been explored so far. We shall argue that explanatory considerations support at best a weak version of the argument from the cognitive science practice, according to which semantic properties concur with formal syntactic properties in individuating computations in cognitive science, but not a strong version, according to which computation individuation in cognitive science is semantic as opposed to formal syntactic. (shrink)