The predominance of machine learning based techniques in cognitive neuroscience raises a host of philosophical and methodological concerns. Given the messiness of neural activity, modellers must make choices about how to structure their raw data to make inferences about encoded representations. This leads to a set of standard methodological assumptions about when abstraction is appropriate in neuroscientific practice. Yet, when made uncritically these choices threaten to bias conclusions about phenomena drawn from data. Contact between the practices of multivariate pattern analysis (...) and philosophy of science can help to illuminate the conditions under which we can use artificial neural networks to better understand neural mechanisms. This paper considers a specific technique for MVPA called representational similarity analysis. I develop a theoretically-informed account of RSA that draws on early connectionist research and work on idealization in the philosophy of science. By bringing a philosophical account of cognitive modelling in conversation with RSA, this paper clarifies the practices of neuroscientists and provides a generalizable framework for using artificial neural networks to study neural mechanisms in the brain. (shrink)

Are there representations in the brain? It depends on what you mean by representations, and it depends on what you want them to do for you—both in terms of the causal role they play in the system, and in terms of their explanatory value. But ideally, we would like an account of representation that allows us to assign a representational role and content to the appropriate mechanistic precursors of behavior that in fact play that role and conversely, search for the (...) mechanistic realizers of representational roles that are posited by our models of behavior and cognition. Such an account would be methodologically valuable in neuroscience. Lately, people have started asking similar questions about deep neural networks. What representations do they learn and use, and what is the relation between those representations and the sometimes impressive capacities these networks exhibit? More philosophically puzzling, perhaps, is the question of why the internal activities or dispositions labeled as such in neuroscience and AI research should count as representations at all. I think we can give a unified answer to both sets of questions, in the form of a kind of representational pragmatism. What makes something a representation just is that we can identify and re-identify it as such, and moreover, that we can manipulate it effectively to do whatever it is that we think representations ought to do for us in that particular context. For the first condition, the idea is that so long as we have some probe that allows us to pick out the relevant causally effective candidate, we should consider that to be a candidate representation-relative-to-that-probe. The second condition can be understood as a kind of anti-gerrymandering constraint: one ought to be able to intervene on the candidate, and see effects of that intervention consistent with the functional role the representation is supposed to play. Naturally the details need to be spelled out for different contexts—and I will articulate them for a few illustrative cases—which effectively allows for a kind of pluralism about representation depending on the setting and the kinds of questions being asked. With the extra components filled in, representational pragmatism can make sense of existing practices in neuroscience and AI, as well as their relationship to naturalistic theories of representation in philosophy. (shrink)

Models developed using machine learning are increasingly prevalent in scientific research. At the same time, these models are notoriously opaque. Explainable AI aims to mitigate the impact of opacity by rendering opaque models transparent. More than being just the solution to a problem, however, Explainable AI can also play an invaluable role in scientific exploration. This paper describes how post-hoc analytic techniques from Explainable AI can be used to refine target phenomena in medical science, to identify starting points for future (...) investigations of causal relationships, and to generate possible explanations of target phenomena in cognitive science. In this way, this paper describes how Explainable AI—over and above machine learning itself—contributes to the efficiency and scope of data-driven scientific research. (shrink)

Will brain imaging technology soon enable neuroscientists to read minds? We cannot answer this question without some understanding of the state of the art in neuroimaging. But neither can we answer this question without some understanding of the concept invoked by the term "mind reading." This article is an attempt to develop such understanding. Our analysis proceeds in two stages. In the first stage, we provide a categorical explication of mind reading. The categorical explication articulates empirical conditions that must be (...) satisfied if mind reading is to be achieved. In the second stage, we develop a metric for judging the proficiency of mind reading experiments. The conception of mind reading that emerges helps to reconcile folk psychological judgments about what mind reading must involve with the constraints imposed by empirical strategies for achieving it. (shrink)

In their attempt to connect the workings of the human mind with their neural realizers, cognitive neuroscientists often bracket out individual differences to build a single, abstract model that purportedly represents (almost) every human being’s brain. In this paper I first examine the rationale behind this model, which I call ‘Platonic Brain Model’. Then I argue that it is to be surpassed in favor of multiple models allowing for patterned inter-individual differences. I introduce the debate on legitimate (and illegitimate) ways (...) of mapping neural structures and cognitive functions, endorsing a view according to which function-structure mapping is context-sensitive. Building on the discussion of the ongoing debate on the function(s) of the so-called Fusiform “Face” Area, I show the necessity of indexing function-structure mappings to some populations of subjects, clustered on the basis of factors such as their expertise in a given domain. (shrink)

Multivariate pattern analysis, or MVPA, has become one of the most popular analytic methods in cognitive neuroscience. Since its inception, MVPA has been heralded as offering much more than regular univariate analyses, for—we are told—it not only can tell us which brain regions are engaged while processing particular stimuli, but also which patterns of neural activity represent the categories the stimuli are selected from. We disagree, and in the current paper we offer four conceptual challenges to the use of MVPA (...) to make claims about neural representation. Our view is that the use of MVPA to make claims about neural representation is problematic. (shrink)

Functional neuroimaging is sometimes criticized as showing only where in the brain things happen, not how they happen, and thus being unable to inform us about questions of mental and neural representation. Novel analytical methods increasingly make clear that imaging can give us access to constructs of interest to psychology. In this paper I argue that neuroimaging can give us an important, if limited, window into the large-scale structure of neural representation. I describe Representational Similarity Analysis, increasingly used in neuroimaging (...) studies, and lay out desiderata for representations in general. In that context I discuss what RSA can and cannot tell us about neural representation. I compare RSA with fMRI to a different experimental paradigm which has been embraced as being indicative of representation in psychology, and argue that it compares favorably. (shrink)

People are motivated by shared social values that, when held with moral conviction, can serve as compelling mandates capable of facilitating support for ideological violence. The current study examined this dark side of morality by identifying specific cognitive and neural mechanisms associated with beliefs about the appropriateness of sociopolitical violence, and determining the extent to which the engagement of these mechanisms was predicted by moral convictions. Participants reported their moral convictions about a variety of sociopolitical issues prior to undergoing functional (...) MRI scanning. During scanning, they were asked to evaluate the appropriateness of violent protests that were ostensibly congruent or incongruent with their views about sociopolitical issues. Complementary univariate and multivariate analytical strategies comparing neural responses to congruent and incongruent violence identified neural mechanisms implicated in processing salience and in the encoding of subjective value. As predicted, neuro-hemodynamic response was modulated parametrically by individuals’ beliefs about the appropriateness of congruent relative to incongruent sociopolitical violence in ventromedial prefrontal cortex, and by moral conviction in ventral striatum. Overall moral conviction was predicted by neural response to congruent relative to incongruent violence in amygdala. Together, these findings indicate that moral conviction about sociopolitical issues serves to increase their subjective value, overriding natural aversion to interpersonal harm. (shrink)

In recent years, neuroscience has begun to transform itself into a “big data” enterprise with the importation of computational and statistical techniques from machine learning and informatics. In addition to their translational applications such as brain-computer interfaces and early diagnosis of neuropathology, these tools promise to advance new solutions to longstanding theoretical quandaries. Here I critically assess whether these promises will pay off, focusing on the application of multivariate pattern analysis (MVPA) to the problem of reverse inference. I argue that (...) MVPA does not inherently provide a new answer to classical worries about reverse inference, and that the method faces pervasive interpretive problems of its own. Further, the epistemic setting of MVPA and other decoding methods contributes to a potentially worrisome shift towards prediction and away from explanation in fundamental neuroscience. (shrink)

In How We Understand Others: Philosophy and Social Cognition, Shannon Spaulding develops a novel account of social cognition with pessimistic implications for mindreading accuracy: according to Spaulding, mistakes in mentalizing are much more common than traditional theories of mindreading commonly assume. In this commentary, I push against Spaulding’s pessimism from two directions. First, I argue that a number of the heuristic mindreading strategies that Spaulding views as especially error prone might be quite reliable in practice. Second, I argue that current (...) methods for measuring mindreading performance are not well-suited for the task of determining whether our mental-state attributions are generally accurate. I conclude that any claims about the accuracy or inaccuracy of mindreading are currently unjustified. (shrink)

In this article, I show the role that the philosopher of cognitive science can cur-rently play in cognitive science research. I argue for the important, and not yet considered, role of the philosophy of cognitive science in cognitive science, that is, the importance of cooperation between philosophers of science with cogni-tive scientists in investigating the research methods and theoretical assump-tions of cognitive science. At the beginning of the paper I point out, how the philosopher of science, here, the philosopher of (...) cognitive science, can participate in interdisciplinary research. I am opting of the cooperation in investigating the so-called reflective problems. Then, I discuss four examples of issues important for the cognitive science, in which the competences possessed by the philosopher are useful. At the ending I point out wider landscape of possible cooperation of philosophers with cognitive scientists. (shrink)

Autonomist accounts of cognitive science suggest that cognitive model building and theory construction (can or should) proceed independently of findings in neuroscience. Common functionalist justifications of autonomy rely on there being relatively few constraints between neural structure and cognitive function (e.g., Weiskopf, 2011). In contrast, an integrative mechanistic perspective stresses the mutual constraining of structure and function (e.g., Piccinini & Craver, 2011; Povich, 2015). In this paper, I show how model-based cognitive neuroscience (MBCN) epitomizes the integrative mechanistic perspective and concentrates (...) the most revolutionary elements of the cognitive neuroscience revolution (Boone & Piccinini, 2016). I also show how the prominent subset account of functional realization supports the integrative mechanistic perspective I take on MBCN and use it to clarify the intralevel and interlevel components of integration. (shrink)

Barsalou has recently argued against the strategy of identifying amodal neural representations by using their cross-modal responses (i.e., their responses to stimuli from different modalities). I agree that there are indeed modal structures that satisfy this “cross-modal response” criterion (CM), such as distributed and conjunctive modal representations. However, I argue that we can distinguish between modal and amodal structures by looking into differences in their cross-modal responses. A component of a distributed cell assembly can be considered unimodal because its responses (...) to stimuli from a given modality are stable, whereas its responses to stimuli from any other modality are not (i.e., these are lost within a short time, plausibly as a result of cell assembly dynamics). In turn, conjunctive modal representations, such as superior colliculus cells in charge of sensory integration, are multimodal because they have a stable response to stimuli from different modalities. Finally, some prefrontal cells constitute amodal representations because they exhibit what has been called ‘adaptive coding’. This implies that their responses to stimuli from any given modality can be lost when the context and task conditions are modified. We cannot assign them a modality because they have no stable relation with any input type. (shrink)

We argue that one important aspect of the “cognitive neuroscience revolution” identified by Boone and Piccinini :1509–1534. doi: 10.1007/s11229-015-0783-4, 2015) is a dramatic shift away from thinking of cognitive representations as arbitrary symbols towards thinking of them as icons that replicate structural characteristics of their targets. We argue that this shift has been driven both “from below” and “from above”—that is, from a greater appreciation of what mechanistic explanation of information-processing systems involves, and from a greater appreciation of the problems (...) solved by bio-cognitive systems, chiefly regulation and prediction. We illustrate these arguments by reference to examples from cognitive neuroscience, principally representational similarity analysis and the emergence of dynamical models as a central postulate in neurocognitive research. (shrink)

This paper concerns local yet systematic problems of contrastive underdetermination of model choice in cognitive neuroscience debates about the so-called two visual systems hypothesis. The underdetermination problem is systematically generated by the way certain assumptions about the representationalist nature of computation are translated into experimental practice. The problem is that behavioural data underdetermine the choice between competing representational models. In this paper, I diagnose how these assumptions generate underdetermination problems in the choice between competing functional models of perception–action. Using the (...) tools of philosophy of science, I describe the type of underdetermination and sketch a possible cure. (shrink)

Since the 90’s, neurolaw is on the rise. At the heart of heated debates lies the recurrent theme of a neuro-revolution of criminal responsibility. However, caution should be observed: the alleged foundations of criminal responsibility (amongst which free will) are often inaccurate and the relative imperviousness of its real foundations to scientific facts often underestimated. Neuroscientific findings may impact on social institutions, but only insofar as they also engage in a political justification of the changes being called for, convince populations, (...) and take into consideration the ensuing consequences. Moreover, the many limits of neuroscientific tools call for increased vigilance when, if ever, using neuroscientific evidence in a courtroom. In this article, we aim at setting the basis for future sound debates on the contribution of neuroscience to criminal law, and in particular to the assessment of criminal responsibility. As such, we provide analytical tools to grasp the political and normative nature of criminal responsibility and review the current or projected use of neuroscience in the law, all the while bearing in mind the highly-publicized question: can neuroscience revolutionize criminal responsibility? Answering this question implicitly requires answering a second question: should neuroscience revolutionize the institution of criminal responsibility? Answering both, in turn, requires drawing the line between science and normativity, revolution and dialogue, fantasies and legitimate hopes. (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)

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...

Will brain imaging technology soon enable neuroscientists to read minds? We cannot answer this question without some understanding of the state of the art in neuroimaging. But neither can we answer this question without some understanding of the concept invoked by the term “mind reading.” This article is an attempt to develop such understanding. Our analysis proceeds in two stages. In the first stage, we provide a categorical explication of mind reading. The categorical explication articulates empirical conditions that must be (...) satisfied if mind reading is to be achieved. In the second stage, we develop a metric for judging the proficiency of mind reading experiments. The conception of mind reading that emerges helps to reconcile folk psychological judgments about what mind reading must involve with the constraints imposed by empirical strategies for achieving it. (shrink)