Abstract
I argue that there are different orders of mechanisms with different constitutive relevance and individuation conditions. In common first-order mechanistic explanations, constitutive relevance norms are captured by the matched-interlevel-experiments condition (Craver et al. (2021) Synthese 199:8807–8828). Regarding individuation, we say that any two mechanisms are of the same type when they have the same concrete components performing the same activities in the same arrangement. By contrast, in higher-order mechanistic explanations, we formulate the decompositions in terms of generalized basic components (GBCs). These GBCs (e.g., logic gates) possess causal properties that are common to a set of physical systems. Mechanistic explanations formulated in terms of GBCs embody the epistemic value of horizontal integration, which aims to explain as many phenomena as possible with a minimal amount of abstract components (Wajnerman Paz (2017a) Philos Psychol 30:213–234). Two higher-order mechanisms are of the same type when they share all the same GBCs, and they are organized as performing the same activities with the same interactions. I use this notion of mechanistic order to enhance the mechanistic account of computation (MAC) and provide an account of the epistemic norms of computational explanation and the nature of medium-independent functional properties and mechanisms. Finally, I use these new conceptual tools to address four criticisms of the MAC.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Wajnerman Paz (2017a) explains: “… the kind of unification that I am considering in this section occurs when different phenomena can be explained by different models that describe sets of shared or common basic components or operations”. I use this idea to formulate an explicit constitutive relevance condition for abstract mechanisms.
I will call GBC the criterion and GBCs (plural form) the abstract components used to formulate an abstract mechanistic explanation. Context should make clear which one is being referred to in every case.
Coelho Mollo provides his own solution to this problem. However, his solution is based on assuming a distinction between mechanistic implementation and functional individuation, which entails that there are no higher-order mechanisms in the sense defined in this article. I will reject this solution since, as I will argue, abstract mechanisms are genuine mechanisms, and I will offer an alternative solution.
Kuokkanen speaks of descriptive abstraction and claims the MAC does not entail an ontological commitment to platonic abstract objects. While I agree with the latter, I disagree that a merely descriptive view of abstraction is needed to avoid such a commitment. We are presented here with a false dilemma. Consider the following example. When describing something as a bacterium, many details about it are omitted and, therefore, to say “X is a bacterium” is, in this sense, an abstract description—as many of its components are replaced over time without affecting its identity as an individual organism. However, we would hardly say that “bacterium” is either a platonic abstract object or a mere descriptive abstraction, as this would imply that even you and I are abstract objects. Consequently, there is a sense in which abstraction is constitutive of what an individual is, given its individuation conditions. For this reason, I will speak of an additional constitutive criterion rather than a descriptive one.
This is how I interpret his article based on the following: (1) he refers to the account in these terms: “the arguments from explanatory value and multiple realizability are not real problems for the MAC as a single hierarchy view” (p. 370), and (2) the fourth section is entitled: “The non-problems and the problem of MAC as a single hierarchy view”.
The notion of “implementation mechanism” is to be interpreted in the following manner: (1) they are the explanandum of a mechanistic explanation that includes more than mathematical and topological structure details, and (2) in addition, the target system of that explanation should be susceptible of a computational (medium-independent) mechanistic explanation—i.e., this target system’s behavior has at least two possible mechanistic explanations—corresponding to the putative computation it implements.
Piccinini (2020, pp. 36–37) advances an egalitarian ontology of levels to deal with causal exclusion. According to him, higher-level properties are aspects of their realizer properties that are invariant under some changes.
The reason is that with respect to motifs, Levy (2016) offers only the following conditional claim: “if, and to the extent that motif-based analysis proves important and informative, it will be a distinctively non-level-like way of thinking about the brain”.
Here ‘neuron’ is not characterized in terms of specific biophysical properties but in a medium medium-independent way.
Algorithms are usually defined as effective procedures to compute the values of specific functions. For example, the algorithm employed to multiply two numbers that have more than one digit each, or, the algorithm used to obtain the least common multiple, which is usually taught in school.
Higher orders are obtained by horizontal abstraction. However, what I call the highest order is technically not an order properly since it is not formulated in terms of a mechanism using GBCs. Nonetheless, I have decided to call it that way by analogy with the orders since it corresponds to the maximum possible level of horizontal abstraction.
Although they can be part of mechanistic explanations, as explanandum phenomena.
In the above example, if we were to abstract away the fact that the music box produces sounds and preserve only the mathematical structure of those sounds, we would not retain its teleofunctional identity as a music box. More precisely, its identity as a mechanism whose teleological function is to produce sounds from input patterns would be lost.
One can still argue that, while the requirement is impossible to meet, it could act as a regulative epistemic ideal, in the sense that it would be better to have no abstraction at all (thanks to an anonymous reviewer for making this point). However, this would conflict with the generality of the explanations in science, and, for that reason, would be irrational for the mechanist to maintain such an epistemic norm. In the most extreme case, horizontal integration and the GBC norm for higher-order mechanisms, we go exactly in the opposite direction.
The reason is not that “dimensions of variation” is intrinsically wrong, but that it could lead to the kind of misunderstanding accounted for here. If, by contrast, dimensions of variations are interpreted as pertaining to degrees of freedom (such as digits)—digits are individuated only by the capacity of the physical system to distinguish a finite number of states from each other, and no reference to a physical variable is needed—rather than specific physical variables (such as shape), the problem vanishes (Piccinini, personal communication).
References
Anderson, N. G., & Piccinini, G. (forthcoming). The physical signature of computation: A robust mapping account. Oxford University Press.
Bechtel, W. (2009). Looking down, around, and up: Mechanistic explanation in psychology. Philosophical Psychology, 22(5), 543–564. https://doi.org/10.1080/09515080903238948
Boone, W., & Piccinini, G. (2016). Mechanistic Abstraction. Philosophy of Science, 83(5), 686–697. https://doi.org/10.1086/687855
Carandini, M., & Heeger, D. J. (2012). Normalization as a canonical neural computation. Nature Reviews Neuroscience, 13(1), 51–62. https://doi.org/10.1038/nrn3136
Carrillo, N., & Knuuttila, T. (2023). Mechanisms and the problem of abstract models. European Journal for Philosophy of Science, 13(3), 27. https://doi.org/10.1007/s13194-023-00530-z
Chalmers, D. J. (1996). Does a rock implement every finite-state automaton? Synthese, 108(3), 309–333. https://doi.org/10.1007/BF00413692
Chirimuuta, M. (2014). Minimal models and canonical neural computations: The distinctness of computational explanation in neuroscience. Synthese, 191(2), 127–153. https://doi.org/10.1007/s11229-013-0369-y
Chirimuuta, M. (2018). Explanation in Computational Neuroscience: Causal and Non-causal. The British Journal for the Philosophy of Science, 69(3), 849–880. https://doi.org/10.1093/bjps/axw034
Coelho Mollo, D. (2018). Functional individuation, mechanistic implementation: The proper way of seeing the mechanistic view of concrete computation. Synthese, 195(8), 3477–3497. https://doi.org/10.1007/s11229-017-1380-5
Coelho Mollo, D. (2019). Are there teleological functions to compute? Philosophy of Science, 86(3), 431–452. https://doi.org/10.1086/703554
Craver, C. F. (2007). Explaining the brain. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199299317.001.0001
Craver, C. F., Glennan, S., & Povich, M. (2021). Constitutive relevance & mutual manipulability revisited. Synthese, 199(3–4), 8807–8828. https://doi.org/10.1007/s11229-021-03183-8
Cummins, R. (1989). Meaning and mental representation. MIT Press.
Elber-Dorozko, L., & Shagrir, O. (2021). Integrating computation into the mechanistic hierarchy in the cognitive and neural sciences. Synthese, 199(S1), 43–66. https://doi.org/10.1007/s11229-019-02230-9
Fresco, N. (2010). Explaining computation without semantics: Keeping it simple. Minds and Machines, 20(2), 165–181. https://doi.org/10.1007/s11023-010-9199-6
Fresco, N. (2014). Physical computation and cognitive science (1st ed.). Springer.
Fuentes, J. I. (2023). Efficient mechanisms. Philosophical Psychology. https://doi.org/10.1080/09515089.2023.2193216
Garson, J. (2013). The functional sense of mechanism. Philosophy of Science, 80, 317–333. https://doi.org/10.1086/671173
Glennan, S. (2017). The new mechanical philosophy (1st ed.). Oxford University Press.
Haimovici, S. (2013). A problem for the mechanistic account of computation. Journal of Cognitive Science, 14(2), 151–181.
Kaplan, D. M. (2011). Explanation and description in computational neuroscience. Synthese, 183(3), 339–373. https://doi.org/10.1007/s11229-011-9970-0
Kim, J. (1998). Mind in a physical world. MIT Press.
Kitcher, P. (1981). Explanatory unification. Philosophy of Science, 48(4), 507–531. https://doi.org/10.1086/289019
Kuokkanen, J. (2022a). No computation without implementation? A potential problem for the single hierarchy view of physical computation. Synthese, 200(5), 370. https://doi.org/10.1007/s11229-022-03696-w
Kuokkanen, J. (2022b). Vertical-horizontal distinction in resolving the abstraction, hierarchy, and generality problems of the mechanistic account of physical computation. Synthese, 200(3), 247. https://doi.org/10.1007/s11229-022-03725-8
Levy, A. (2016). The unity of neuroscience: A flat view. Synthese, 193(12), 3843–3863. https://doi.org/10.1007/s11229-016-1256-0
Levy, A., & Bechtel, W. (2013). Abstraction and the organization of mechanisms. Philosophy of Science, 80(2), 241–261. https://doi.org/10.1086/670300
Maley, C. J. (2023). Medium independence and the failure of the mechanistic account of computation. Ergo an Open Access Journal of Philosophy, 10, 28. https://doi.org/10.3998/ergo.4658
Marr, D. (1982/2010). Vision: A computational investigation into the human representation and processing of visual information. MIT Press.
Milkowski, M. (2013). Explaining the computational mind. MIT Press.
Piccinini, G. (2015). Physical computation: A mechanistic account. Oxford University Press.
Piccinini, G. (2020). Neurocognitive mechanisms: Explaining biological cognition (1st ed.). Oxford University Press.
Piccinini, G., & Bahar, S. (2013). Neural computation and the computational theory of cognition: Cognitive science. Cognitive Science, 37(3), 453–488. https://doi.org/10.1111/cogs.12012
Piccinini, G., & Craver, C. (2011). Integrating psychology and neuroscience: Functional analyses as mechanism sketches. Synthese, 183(3), 283–311. https://doi.org/10.1007/s11229-011-9898-4
Polger, T. W., & Shapiro, L. A. (forthcoming). The puzzling resilience of multiple realization. Minds and Machines.
Priorelli, M., & Stoianov, I. P. (2023). Flexible intentions: An Active Inference theory. Frontiers in Computational Neuroscience, 17, 1128694. https://doi.org/10.3389/fncom.2023.1128694
Searle, J. R. (1992). The rediscovery of the mind. MIT Press.
Shagrir, O. (2022). The nature of physical computation. Oxford University Press.
Shagrir, O., & Bechtel, W. (2017). Marr’s computational level and delineating phenomena. In D. M. Kaplan (Ed.), Explanation and integration in mind and brain science (pp. 190–214). Oxford University Press.
Sterling, P., & Laughlin, S. (2015). Principles of neural design. MIT Press.
Todesco, G. M. (2013). Cellular automata: The game of life. In M. Emmer (Ed.), Imagine math 2: Between culture and mathematics (pp. 231–243). Springer.
Wajnerman Paz, A. (2017a). A mechanistic perspective on canonical neural computation. Philosophical Psychology, 30(3), 213–234. https://doi.org/10.1080/09515089.2016.1271117
Wajnerman Paz, A. (2017b). Pluralistic mechanism. Theoria, 32(2), 161–175. https://doi.org/10.1387/theoria.16877
Acknowledgements
I want to thank the University of Missouri—St. Louis and especially Gualtiero Piccinini for being such a great host and a very insightful commentator of my manuscripts during my research time there. Other people in St Louis were very supportive during my time there, especially those participating in The Philosophers’ Forum (such as Timothy Luft, Zhang Ling, Henrique Cassol Leal, and Logan Bohlinger, among others), David McGraw, and Nicolae Viorel Burcea. Additionally, this research would not have been possible without the guidance and comments made by Abel Wajnerman Paz. I am also thankful to Juan Manuel Garrido for his academic advice whenever it was needed. Two anonymous reviewers from Synthese made significant contributions through their comments. Lastly, I would like to express my most profound appreciation to my wife, Jessica Vásquez, for the support she has given me in developing my career.
Funding
Agencia Nacional de Investigación y Desarrollo (Chile). Beca de Doctorado Nacional 2020/21202030. Fondo Nacional de Desarrollo Científico y Tecnológico (Chile). Fondecyt de Iniciación 11220327.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author has no financial or other interests to declare.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Fuentes, J.I. Computational systems as higher-order mechanisms. Synthese 203, 55 (2024). https://doi.org/10.1007/s11229-023-04482-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11229-023-04482-y