Multivariate pattern analysis and the search for neural representations

Gessell, Bryce; Geib, Benjamin R.; Brigard, Felipe De

doi:10.1007/s11229-021-03358-3

Cited by 8 publications

(7 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For, it is easy to interpret (bona fide) nonrepresentational states in representational terms (Bechtel, 1998;Shapiro, 2013), as well as "deflating" (bona fide) genuine representations as mere causal mediators (Facchin, 2021;Ramsey, 2007)-even when the cognitive system we're looking at is the brain (cf. Gessell et al, 2021;Kriegeskorte & Kievit, 2013;Ritchie et al, 2019).…”

Section: Scientific Cognition and Its Mocmentioning

confidence: 99%

Why can’t we say what cognition is (at least for the time being)

Facchin

2023

PhiMiSci

View full text Add to dashboard Cite

Some philosophers search for the mark of the cognitive: a set of individually necessary and jointly sufficient conditions identifying all instances of cognition. They claim that the mark of the cognitive is needed to steer the development of cognitive science on the right path. Here, I argue that, at least at present, it cannot be provided. First (§2), I identify some of the factors motivating the search for a mark of the cognitive, each yielding a desideratum the mark is supposed to satisfy (§2.1). I then (§2.2) highlight a number of tensions in the literature on the mark of the cognitive, suggesting they’re best resolved by distinguishing two distinct programs. The first program (§3) is that of identifying a mark of the cognitive capturing our everyday notion of cognition. I argue that such a program is bound to fail for a number of reasons: it is not clear whether such an everyday notion exists; and even if it existed, it would not be able to spell out individually necessary and jointly sufficient conditions for cognition; and even if it were able to spell them out, these conditions won’t satisfy the desiderata a mark of the cognitive should satisfy. The second program is that of identifying a mark of the cognitive spelling out a genuine scientific kind. But the current state of fragmentation of cognitive science, and the fact that it is splintered in a myriad of different research traditions, prevent us from identifying such a kind. And we have no reason to think that these various research traditions will converge, allowing us to identify a single mark. Or so, at least, I will argue in (§4). I then conclude the paper (§5) deflecting an intuitive objection, and exploring some of the consequences of the thesis I have defended.

show abstract

Section: Scientific Cognition and Its Mocmentioning

confidence: 99%

Why can’t we say what cognition is (at least for the time being)

Facchin

2023

PhiMiSci

View full text Add to dashboard Cite

show abstract

“…Neuroscientists themselves gesture towards the goal of understanding functional neural mechanisms, but rarely spell out how RSA contributes to this kind of explanation (Kriegeskorte & Diedrichsen, 2019). Hence, explicitly specifying how RSA contributes to mechanistic explanations can help to clarify the framework's utility in the face of recent skepticism about representational explanations afforded by computational models in cognitive neuroscience (Carlson et al, 2018;Ritchie et al 2019;Gessell et al 2021). I aim to do so by drawing on philosophical work on causal patterns, idealization, and connectionist modeling.…”

Section: Models and Their Targets Instantiate Causal Patternsmentioning

confidence: 99%

“…Moreover, by emphasizing mechanistic rather representational explanations, we can somewhat sidestep skeptical worries about the latter. Even in cases where it seems like the specific content of representations is radically underdetermined by the available evidence (Gessell et al, 2021), we can still use RSA to get evidence about the relationship between a model and a target system. This is because our goal is to test whether a model and its target both instantiate some idealized causal pattern.…”

Section: Models and Their Targets Instantiate Causal Patternsmentioning

confidence: 99%

“…Finally, by shifting focus to illuminating mechanisms, we can somewhat sidestep worries about the underdetermination of semantic content by the available evidence (Ritchie et al, 2019;Gessell et al, 2021). Though we are still concerned to understand the content of representations, establishing a high degree of predictive success can be enough to establish that a model instantiates a shared causal pattern.…”

Section: Inferring Neural Mechanisms With Goal-driven Models Via Caus...mentioning

confidence: 99%

See 1 more Smart Citation

Mapping representational mechanisms with deep neural networks

Kieval

2022

Synthese

View full text Add to dashboard Cite

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 (MVPA) 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 (RSA). 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.

show abstract

“…3; Cao, 2022; Baker, Lansdell, & Kording, 2022). Moreover, even an analysis of a multiarea pattern of activation is not sufficient to determine between two different hypotheses about representation (Gessell, Geib, & De Brigard, 2021). Here, we make an even stronger claim: no amount of neuronal data, by itself, can be used to conclusively decide between two competing hypotheses about representation.…”

mentioning

confidence: 99%

Do Retinal Neurons Also Represent Somatosensory Inputs? On Why Neuronal Responses Are Not Sufficient to Determine What Neurons Do

Elber-Dorozko

Loewenstein

2023

Cognitive Science

View full text Add to dashboard Cite

How does neuronal activity give rise to cognitive capacities? To address this question, neuroscientists hypothesize about what neurons “represent,” “encode,” or “compute,” and test these hypotheses empirically. This process is similar to the assessment of hypotheses in other fields of science and as such is subject to the same limitations and difficulties that have been discussed at length by philosophers of science. In this paper, we highlight an additional difficulty in the process of empirical assessment of hypotheses that is unique to the cognitive sciences. We argue that, unlike in other scientific fields, comparing hypotheses according to the extent to which they explain or predict empirical data can lead to absurd results. Other considerations, which are perhaps more subjective, must be taken into account. We focus on one such consideration, which is the purposeful function of the neurons as part of a biological system. We believe that progress in neuroscience critically depends on properly addressing this difficulty.

show abstract

Multivariate pattern analysis and the search for neural representations

Cited by 8 publications

References 41 publications

Why can’t we say what cognition is (at least for the time being)

Why can’t we say what cognition is (at least for the time being)

Mapping representational mechanisms with deep neural networks

Do Retinal Neurons Also Represent Somatosensory Inputs? On Why Neuronal Responses Are Not Sufficient to Determine What Neurons Do

Contact Info

Product

Resources

About