Philosophy of Cognitive Neuroscience: Content, Self, and Cognitive Ontology
37th Annual Boulder Conference on the History and Philosophy of Science
September 9–11, 2022, University of Colorado, Boulder
Keynote Speakers: Adina Roskies (Dartmouth),
Rosa Cao (Stanford), and Michael Anderson (Western U., Ontario)
Advances in cognitive neuroscience promise to shed light on matters of longstanding philosophical interest, including the nature of the self and the contents of thought and perception. At the same time, cognitive neuroscience, as a discipline, is the subject of much philosophical interest. What notion of mental representation, if any, does successful cognitive neuroscience presuppose? What methodological difficulties must be overcome if cognitive neuroscience is to flourish? What is the function of a localized structure or mechanism in the brain? Moreover, cognitive neuroscience provides fertile ground for the exploration of more general questions about methodology and scientific processes, for example, to do with explanation or causal intervention. This conference will provide a venue for the interdisciplinary discussion of all of these issues and more.
Friday, Sept. 9
- 6:00–7:30 p.m. Duane Physics G125
- Keynote: “Neuropsychiatry and the Ontology of Agency,” Adina Roskies (Dartmouth)
- The development of new therapeutic methods for intervening in brain function have the potential to influence a person’s autonomy and authenticity. In order to determine whether or not this is the case, we need an assay that can measure aspects of agency before and after neurointervention. Here I introduce a framework for thinking about agency as a multidimensional construct, and describe a research project aimed at measuring agency. We develop an Agency Assessment Tool (AAT) using survey and behavioral methods. I present preliminary results from the project that show that the AAT is able to discern control and patient groups, and discuss the prospects for using data-driven tools to come up with ontologies of agency.
- Keynote: “Neuropsychiatry and the Ontology of Agency,” Adina Roskies (Dartmouth)
- 7:30–9:00 p.m.
- Reception, 11th-floor lounge, Gamow Tower, Duane Physics Building, hors d’oeuvres
Saturday, Sept. 10
All talks are in Humanities 135
- 9:00–9:50. “A Meta-Informational Theory of Content,” Christopher Viger (Western U., Ontario)
- The brain has a massive interface problem. By the very nature of functional specialization, the detailed information processed by one subsystem is not the kind of information another subsystem is specialized to process; thus, information cannot be the lingua franca of the cognitive system. Graph-theoretic analysis suggests that specialized processing units interact via long-range connections, distinguishing local and global processing. Thus, content transmitted globally must be formatted for use by non-specialists, a kind of summary report or meta-information about the detailed information processing occurring in specialized brain areas: a true lingua franca. Global processing also correlates with consciousness, so the content of conscious thought is meta-informational. One consequence of the brain's solution to its massive interface problem is that there is a kind of user-illusion about the content of conscious thought. A second consequence is that the language of (conscious) thought is a kind of pidgin.
- 10:00–10:50. “Perspectival Inhibition in Perception,” Mia Accomando (Rutgers U.)
- Does perception represent the location of the self, the body, or neither? I develop a novel answer by shedding light on a neural mechanism known as sensory gating. The basic insight is that perception ignores the location of the self by filtering out of experience the location of the body. The location of the body is represented in the motor system as a range of motor possibilities. When a movement is initiated, the sensory gating mechanism makes a partition in this space of motor possibilities and outlines a trace of the perpetual consequences for which one’s own body is the source. This first-personal representation of the body in the perceptual system marks the location of the self and is ignored in perceptual experience. So perception represents neither the location of the body nor the location of the self, while these representations are indeed implicit in the content of perceptual experience.
- 10:50–11:10. Coffee Break
- 11:10–12:00. “The New Central Executive,” David Barack (U. of Pennsylvania)
- 12:00–1:30 Lunch, Boulder Saturday Market
- 1:30–2:20. “Getting Real about Neural Data Science,” Phillip Hintikka Kieval (U. of Cambridge)
- Recent research suggests that manifolds play an important role in neural computations. These manifolds are continuous, low-dimensional structures embedded in high-dimensional neural activity. Investigators purport to uncover these structures by using data-analytic techniques to reduce the dimensionality of patterns of neural activity and subsequently reveal the underlying dynamics that are functionally relevant to a specific task. However, the practice of uncovering low-dimensional structures with dimensionality reduction involves modeling choices that introduce a range of implicit assumptions which threaten to cloud our analysis. Yet, the theoretical importance of these modeling choices are rarely discusses by the modellers themselves. To what extent do these techniques license realist claims about the existence of neural manifolds and their role in representing and computing information in the brain? I argue that low-dimensional structures uncovered through data-analysis are akin to rational reconstructions of the underlying dimensionality on the basis of what is empirically accessible to us. These structures reflect only our analyses and are strictly speaking non-factive. However, we can evaluate competing analyses and adjudicate between cases of underdetermination by asking which analysis best reconstructs our best theoretical hypotheses concerning the neural task under investigation. This approach, however, demands a much more robust role for theoretical work concerning the nature and structure of neural computations.
- 2:30–3:20. “Toward a Predictive Philosophy of Cognitive Neuroscience,” Marco Nathan (U. of Denver)
- Philosophy of science—cognitive neuroscience included—has been dominated by explanation. Yet, the times they are a-changing. Many research projects are increasingly couching their epistemic aims purely in terms of forecasting accuracy. This “predictive turn” calls for a radical reconceptualization. What would a predictive philosophy of cognitive neuropsychology look like? Following some preliminary considerations regarding the current intellectual landscape, I examine the problem of epistemic opacity and black-boxed explanations, which has been causing a stir in machine learning and explainable AI. The final portion of the essay draws apart two roles for prediction to play within a scientific inquiry, which reveals how my proposal is separate from “predictive processing” and related movements. I wrap things up with some general remarks on how to revive prediction as a self-standing theoretical virtue independent of explanation. The crucial shift, I maintain, lies in revisiting the hallowed relation between causation and correlation.
- 3:20–3:45. Coffee Break
- 3:45–5:15 Keynote: "How to be a pragmatist about representation (without going too far!)," Rosa Cao (Stanford U.)
- Are there representations in the brain? And how can we tell? It depends on what you mean by representations, and it depends on what you want them to do for you. I'll present a view that attempts to make sense of existing practices in neuroscience and AI, as well as their relationship to naturalistic theories of representation in philosophy. Representational pragmatism says that representations are probe-relative states whose putative contents provide distinctive explanatory value for understanding the system. According to this view, a candidate (brain state, activation pattern, or what-have-you) should qualify as a representation whenever we have a probe that allows us to consistently pick it out, and when intervening on the candidate produces just those effects predicted by our representational theory of the system. Which probes we should prefer will depend on the explanatory context, enforcing a kind of disciplined pluralism about representation.
- 6:30–8:30. Conference Dinner at Leaf Restaurant
Sunday, Sept. 11
All talks are in HUMN 135
- 9:00–9:50. “Cognitive Paleoethology of Working Memory,” Colin Allen (U. of Pittsburgh)
- There have been attempts to estimate the working memory capacity of our earliest hominin ancestors (e.g. Read 2008). The inferences involve comparison with chimpanzees, and extrapolations from cranial capacity measured from fossils. I will critique the narrow focus on capacity and draw upon more recent comparative work (Völter et al. 2019) and neuroscientific work (Lorenc et al. 2021) to argue instead that a focus on the larger system in which working memory is situated provides a better path towards strong inferences about cognitive evolution in the earliest Homo species some 2 million years ago.
- 10:00–10:50. “The Uses of fMRI: Abstraction, Implementation, and Choosing between the Two,” Grace Huckins (Stanford U.)
- 10:50–11:10. Coffee Break
- 11:10–12:00. “There Cannot Be a Mechanism-Only Theory of Computation,” Danielle Williams (U. of California, Davis)
- What is it to give a theory of computation in terms of mechanisms alone? It has recently been argued that a mechanistic account can provide a completed theory of computation. I argue that a theory that relies on mechanisms alone is conceptually unable to provide a complete account of what it means for a physical system to compute. This paper places the importance of the implementation question within the methodology of cognitive science by focusing on its implications on Marr’s tri-level approach. A theory of implementation is what connects the formal notion of computation specified at Marr’s algorithmic level with the physical implementing structures specified at the implementation level. An answer to the implementation question connects these two relata. A theory that specifies physical computation in terms of mechanisms alone provides only a partial answer to the implementation question and therefore cannot be a complete theory of computation.
- 12:00–1:30. Lunch. Half-Fast Subs
- 1:30–3:00. Keynote: "Neural reuse, dynamics, and constraints: Getting beyond componential mechanistic explanation of neural function," Michael Anderson (Western U., Ontario)
- In this talk, I will review some of the evidence for neural reuse--a form of neural plasticity whereby existing neural resources are put to many different uses--and use it to motivate an argument that we need to move beyond (although not necessarily abandon) componential mechanistic explanation in the neurosciences. I claim that what is needed to capture the full range of neural plasticity and dynamics is a style of explanation based on the notion of constraints--enabling constraints in particular. I will give examples of neural phenomena that are hard to capture in the mechanistic framework, and show that they are naturally handled by enabling constraints. As this moves us away from faculty psychology, it has some important implications for the ontology of cognition. As final speaker, I will also try to capture some of the lessons of the conference.
- 5:00–7:00. HH/end-of-conference drink
This conference is supported by CU Boulder’s Committee on the History and Philosophy of Science and CU Boulder’s Institute of Cognitive Science.
Transportation: The nearest airport is the Denver International Airport (DEN). There is a convenient bus (AB route) from DEN to Boulder (timetables and fee info here). Boulder is very bike-friendly and pedestrian-friendly! Boulder 'BCycle' has rental stations all over town, most of which include e-bikes. Boulder has many busses that run regularly all over town. Routes and schedules here (note Boulder's system is on the denver.rtd site)
Accommodations: A block of rooms has been reserved at the Millennium Harvest House Hotel Boulder. You can book your room online or by phone. The group rate is $125 per night. Use the Group Code CHPSCONFERENCE. This rate applies only to the Standard Room and is good until August 7th only; please book before that date, if you want the discounted rate. In addition to the Millenium Harvest House, there are many lovely AirBnbs in Boulder, as well as the historic Hotel Boulderado.