Multiple Realizability from a Causal Perspective

Ross, Lauren N.

doi:10.1086/709732

Cited by 22 publications

(4 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Second, as philosophers of biology have argued for the last several decades, many of the properties and abilities of biological systems can be multiply realized, that is, they can be realized through different causal mechanisms (Ross, 2020;Sober, 1999). Thus, the idealizations within one model may not be adequate for its application in a different target system.…”

Section: Introductionmentioning

confidence: 99%

Deep problems with neural network models of human vision

Bowers¹,

Dujmović²,

Montero³

et al. 2022

Behav Brain Sci

View full text Add to dashboard Cite

Deep neural networks (DNNs) have had extraordinary successes in classifying photographic images of objects and are often described as the best models of biological vision. This conclusion is largely based on three sets of findings: (1) DNNs are more accurate than any other model in classifying images taken from various datasets, (2) DNNs do the best job in predicting the pattern of human errors in classifying objects taken from various behavioral datasets, and (3) DNNs do the best job in predicting brain signals in response to images taken from various brain datasets (e.g., single cell responses or fMRI data). However, these behavioral and brain datasets do not test hypotheses regarding what features are contributing to good predictions and we show that the predictions may be mediated by DNNs that share little overlap with biological vision. More problematically, we show that DNNs account for almost no results from psychological research. This contradicts the common claim that DNNs are good, let alone the best, models of human object recognition. We argue that theorists interested in developing biologically plausible models of human vision need to direct their attention to explaining psychological findings. More generally, theorists need to build models that explain the results of experiments that manipulate independent variables designed to test hypotheses rather than compete on making the best predictions. We conclude by briefly summarizing various promising modelling approaches that focus on psychological data.

show abstract

Section: Introductionmentioning

confidence: 99%

Deep problems with neural network models of human vision

Bowers¹,

Dujmović²,

Montero³

et al. 2022

Behav Brain Sci

View full text Add to dashboard Cite

show abstract

“…Scientific instances of such situations and the appropriate agnostic response are easy to find, for example, in cases of causal heterogeneity and multiple realizability (Ross, 2020(Ross, , 2022. To focus on one example in particular, here and throughout this paper, Parkinson's disease (PD) is produced by a complex of causal pathways across different patients.…”

Section: Uniqueness and Ibe's Criticsmentioning

confidence: 99%

On the Logical Structure of Best Explanations

Schupbach

2023

Philos. sci.

View full text Add to dashboard Cite

Standard articulations of Inference to the Best Explanation (IBE) imply the uniqueness claim that exactly one explanation should be inferred in response to an explanandum. This claim has been challenged as being both too strong (sometimes agnosticism between candidate explanatory hypotheses seems the rational conclusion) and too weak (in cases where multiple hypotheses might sensibly be conjointly inferred). I propose a novel interpretation of IBE that retains the uniqueness claim while also allowing for agnostic and conjunctive conclusions. I then argue that a particular probabilistic explication of explanatory goodness helpfully guides us in navigating such options when using IBE.

show abstract

“…Secondly, as philosophers of biology have argued for the last decades, many of the properties and abilities of biological systems can be multiply realized, i.e. they can be realized through different causal mechanisms (Sober 1999;Ross 2020). Thus, the idealizations within one model may not be adequate for its application in a different target system.…”

Section: Main Textmentioning

confidence: 99%

Neural networks, AI, and the goals of modeling

Veit,

Browning

2023

Behav Brain Sci

View full text Add to dashboard Cite

Deep neural networks (DNNs) have found many useful applications in recent years. Of particular interest have been those instances where their successes imitate human cognition and many consider artificial intelligences to offer a lens for understanding human intelligence. Here, we criticize the underlying conflation between the predictive and explanatory power of DNNs by examining the goals of modeling.

show abstract

Multiple Realizability from a Causal Perspective

Cited by 22 publications

References 37 publications

Deep problems with neural network models of human vision

Deep problems with neural network models of human vision

On the Logical Structure of Best Explanations

Neural networks, AI, and the goals of modeling

Contact Info

Product

Resources

About