| 书目名称 | Neural Machines: A Defense of Non-Representationalism in Cognitive Neuroscience | | 编辑 | Matej Kohár | | 视频video | | | 概述 | Presents a novel defense of non-representationalism.Synthesizes mechanistic explanation with non-representationalism.Furthers the debate between representationalism and non-representationalism | | 丛书名称 | Studies in Brain and Mind | | 图书封面 |  | | 描述 | .In this book, Matej Kohar demonstrates how the new mechanistic account of explanation can be used to support a non-representationalist view of explanations in cognitive neuroscience, and therefore can bring new conceptual tools to the non-representationalist arsenal. Kohar focuses on the explanatory relevance of representational content in constitutive mechanistic explanations typical in cognitive neuroscience. The work significantly contributes to two areas of literature: 1) the debate between representationalism and non-representationalism, and 2) the literature on mechanistic explanation..Kohar begins with an introduction to the mechanistic theory of explanation, focusing on the analysis of mechanistic constitution as the basis of explanatory relevance in constitutive mechanistic explanation. He argues that any viable analysis of representational contents implies that content is not constitutively relevant to cognitive phenomena. The author also addresses objections against his argument and concludes with an examination of the consequences of his account for both traditional cognitive neuroscience and non-representationalist alternatives. This book is of interest to readers in | | 出版日期 | Book 2023 | | 关键词 | Mechanistic Explanation; Cognitive Neuroscience; Anti-representationalism in Cognitive Neuroscience; Ex | | 版次 | 1 | | doi | https://doi.org/10.1007/978-3-031-26746-8 | | isbn_softcover | 978-3-031-26748-2 | | isbn_ebook | 978-3-031-26746-8Series ISSN 1573-4536 Series E-ISSN 2468-399X | | issn_series | 1573-4536 | | copyright | The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerl |
| 1 |
Front Matter |
|
|
Abstract
|
| 2 |
,Introduction, |
Matej Kohár |
|
Abstract
The goal of cognitive neuroscience is to uncover neural mechanisms responsible for intelligent behaviour in humans and animals. Intelligent behaviour has traditionally been taken to include decision-making, use of language and other high-level cognitive phenomena. Over time, however, the scope of what is meant by intelligent behaviour for the purposes of determining the proper subject matter of cognitive sciences (including cognitive neuroscience) has expanded to include any context-dependent responses to stimuli. Cognitive neuroscience therefore engages in search for neural mechanisms underlying sensory perception, memory, navigation, object-recognition, tracking, avoidance, etc. That is, the scope of cognitive neuroscience covers the search for neural mechanisms all the way from sensory processing, through response selection to motor control. Importantly, the scope of the field is not confined to a single species, such as the human, but includes, at least in principle, also the study of animal cognition – either for its own sake or as a model for the human case, when ethical and/or practical considerations prohibit investigation into the human case directly.
|
| 3 |
,The New Mechanistic Theory of Explanation: A Primer, |
Matej Kohár |
|
Abstract
In this chapter, I introduce the new mechanistic framework of explanation, focusing in particular on constitutive mechanistic explanation. I evaluate the accounts of mechanistic constitution derived from Craver’s influential mutual manipulability account. I argue that the horizontal surgicality account due to Baumgartner, Casini and Krickel is compatible with the reciprocal manipulability account due to Krickel, and that together they can be used to evaluate claims about mechanistic constitution. Furthermore, I review the established accounts of how mechanistic models are developed and compare the mechanistic framework of explanation with other theories of scientific explanation, such as the covering-law model, unificationism and older causal theories of explanation.
|
| 4 |
,Mechanistic Explanatory Texts, |
Matej Kohár |
|
Abstract
In this chapter, I address the question how mechanistic models provide why-explanations. I argue for a contrastive approach to mechanistic explanation along the lines of Craver and Kaplan. However, I show that Craver and Kaplan’s view leads to four problems: the Contrasts in Ontology problem, the problem of Identifying Relevant Switch-Points, the problem of Single-Model Explanation and the problem of Empirical Adequacy. I show that the problems can be resolved by adopting a distinction between ontic mechanisms, mechanism descriptions and mechanistic explanatory texts. Mechanistic explanatory texts are answers to contrastive why-questions formulated by comparing the actual mechanism for a phenomenon with a counterfactual mechanism which would underlie the closest contrast phenomenon.
|
| 5 |
,Representations and Mechanisms Do Not Mix, |
Matej Kohár |
|
Abstract
In this chapter, I outline an argument against explanatory relevance of neural representations. After clarifying the concept of neural representation and distinguishing it from person-level mental representation, I argue that a truly representational explanation requires that the content of neural representations be explanatorily relevant. I then argue that this is inconsistent with the requirements for mechanistic constitution, and consequently a constitutive mechanistic explanation cannot be representational. The inconsistency stems from the fact that naturalistic analyses of representational contents render them either non-local to, or not mutually dependent with cognitive phenomena (the details of this argument are expounded in Chaps. ., ., and .). This argument is then compared with its forerunners in the literature – the arguments about causal exclusion of contents; the arguments about internalism/externalism about meanings and mental state individuation; and from methodological solipsism. I also defend this argument from prima facie objections.
|
| 6 |
,Indicator Contents, |
Matej Kohár |
|
Abstract
In this chapter I argue that indicator contents cannot be explanatorily relevant in constitutive mechanistic explanations of cognitive phenomena. The argument relies on the observation that indicator contents are based in conditional probabilities. I first argue that these probabilities must be viewed as physical chances rather than epistemic probabilities or credences. Then I examine the most prominent views on the nature of physical chances to figure out which states of affairs are described by the claim that a neural vehicle carries a particular content. I show that, under both frequentist and propensity interpretations of chances, the property of carrying representational content X is not local to any cognitive phenomena. Furthermore, I show that this property is only mutually dependent with cognitive phenomena under a long-run propensity interpretation of physical chances. However, owing to the failure of locality, I conclude that indicator contents are never explanatorily relevant in constitutive mechanistic explanations in neuroscience.
|
| 7 |
,Structural Contents, |
Matej Kohár |
|
Abstract
In this chapter, I argue that contents of structural representations cannot be explanatorily relevant in constitutive mechanistic explanations of cognitive phenomena. After introducing structural representations, I argue that while there are superficial differences in structuralist theories of representational content, the paradigm cases of structural representation are best understood if the content-determining relation is viewed liberally as simple second-order resemblance. I then argue that contents based on second-order resemblance are not local to cognitive phenomena, because, paradigmatically, second-order resemblance is a relation between a neural assembly and the environment. This is especially pronounced in cases of surrogative reasoning, where the represented objects are absent, and in cases of misrepresentation, where the representing neural assembly does not in fact resemble the target domain. On the other hand, I argue that contents of structural representations are mutually dependent with cognitive phenomena. I also discuss the special case of neural emulators representing other brain areas or other parts of the organism. I conclude that these may be explanatorily rel
|
| 8 |
,Teleosemantics, |
Matej Kohár |
|
Abstract
In this chapter, I argue that representational contents based on teleosemantics cannot be explanatorily relevant in constitutive mechanistic explanations of cognitive phenomena. I note that most teleosemantic theories of content rely on indicator relations or structural similarity in addition to teleofunctions. As such, they inherit the issues with constitutive relevance inherent to indicator semantics and structuralist semantics. I furthermore show that bringing teleofunctions into the mix makes the situation worse. In order to evaluate whether teleofunctional contents are constitutively relevant, I show what carrying a particular representational content comes down to on a teleosemantic account. I argue that this depends on one’s preferred analysis of teleofunctions. I consider etiological functions, historical functions, modal functions, as well as synchronic and cybernetic analyses of function. I argue that all of the surveyed analyses of function render representational contents non-local to cognitive phenomena. The mutual dependence requirement is only fulfilled if the cybernetic analysis is used. Hence, teleofunctional contents are not constitutively relevant. The only possi
|
| 9 |
,The Dual-Explananda Defence, |
Matej Kohár |
|
Abstract
In this chapter, I discuss the view that complete explanations of cognitive phenomena require accounting for explananda which cannot be explained purely mechanistically and must be explained representationally instead. I term this view the dual-explananda defence. I identify two types of dual explananda: (1) fittingness – the fact that particular mechanisms are well suited for particular cognitive tasks and not others; and (2) success: the fact that an organism is successful or unsuccessful in performing a cognitive task. I argue that the mechanistic framework can robustly account for these explananda with the help of the conceptual tools developed in Chap. . – contrastive formulations of explananda, and the use of comparisons to formulate mechanistic explanatory texts.
|
| 10 |
,The Pragmatic Necessity Defence, |
Matej Kohár |
|
Abstract
In this chapter, I discuss the view that referring to representational contents is necessary in non-explanatory/extratheoretical contexts. I concentrate on Egan’s account of the cognitive/intentional gloss and its functions. On my reading, the functions of the intentional gloss include connecting scientific theory with pretheoretical explananda, acting as a heuristic, tracking the progress of neural computations, and playing a didactic role. I argue that the use of representational contents to connect cognitive theory with pre-theoretical explananda obscures a series of complex mappings between common-sense concepts, scientific concepts, tasks, and neural activity, which is made all the more relevant by the increasing evidence of neural reuse. I further argue that the use of representational contents for didactic purposes may have pernicious effects for this very reason. I accept that representational contents might have a heuristic and tracking role in the context of discovery, but contend that there are other usable heuristics, and should representational contents be used in this way, care must be taken to clearly differentiate between the context of discovery and the context of
|
| 11 |
,Conclusions and Future Directions, |
Matej Kohár |
|
Abstract
In this chapter, I highlight the consequences of my central thesis – that representational contents cannot be explanatorily relevant in mechanistic explanations – for both mainstream cognitive neuroscience and for pre-existing non-representational alternatives. I argue that for mainstream cognitive neuroscience, this book has shown representations to be of little importance, and that mechanistic explanatory texts should not mention representational contents. I further argue that the non-representational mechanistic approach enables better understanding of some experimental practices in cognitive neuroscience. The consequences for pre-existing non-representational approaches consist mainly in enabling hybrid dynamical/mechanistic and enactive/mechanistic explanations. I discuss in more detail how dynamical descriptions of mechanism components may be used in mechanistic explanatory texts. Finally, I discuss two avenues for future research – the possibility of using mechanistic compositionality to resolve the scaling-up/representation-hunger problem and the possibility of generalising the account of mechanistic non-representational explanation from neuroscience to psychology.
|
| 12 |
Back Matter |
|
|
Abstract
|
|
|
发表于 2025-3-21 18:02:07
