Online Seminar Series: Problem of Cognitive Ontology
Functional brain imaging studies try to map patterns of activation to cognitive functions, and usually rely upon functional task decomposition based on hypotheses derived from intuition and cognitive psychology. The tasks we postulate constitute a cognitive ontology. What is the epistemic status of these functional commitments? Do we have reason to believe they accurately track the fundamental building blocks of cognition? Does the idea that cognition has fundamental building blocks have merit? Is there a way to bootstrap ourselves out of mistaken theories, or are the methods of neuroimaging ill-suited to alert us to mistaken views? Can other areas of neuroscience help constrain our ontologies? What is the upshot of the issues for gaining knowledge from neuroimaging studies and for theories of scientific realism more generally?
This interdisciplinary series will focus upon these questions and their philosophical implications, and will explore possible methods for addressing these philosophical concerns, such as data-driven discovery methods for cognitive functions.
Talks will be presented online throughout the fall 2020 semester, schedule is below.
Register using this link: https://pitt.zoom.us/webinar/register/WN_KMNKu4fmQ9Wh5ZjvXJ3qQA
Please note, registration will be for the entire seminar series.
Mondays 1:30 – 3 pm Eastern Time Zone
Sept 21: Russ Poldrack (Stanford University), Cognitive Ontologies, from Top to Bottom
ABSTRACT: Cognitive ontologies have been primarily developed in a top-down manner on the basis of psychological theories that lean heavily upon folk psychological concepts. I will first assess the degree to which these ontologies are effective in predicting brain activity using cognitive encoding models, showing that they are surprisingly effective but also highlighting some important shortcomings. I will then ask whether we might develop better understanding of the organization of the mind through bottom-up, or “data driven” ontologies. I will present two examples of such data-driven ontologies, outlining their success but also detailing important limitations on this approach.
Uljana Feest (Leibniz Universität Hannover), “Cognitive Kinds and Investigative Practice”
ABSTRACT: When psychologists investigate their objects of research, such as (kinds of) memory, they operationally define these “objects” in terms of particular experimental tests/tasks, assumed to provide epistemic access to the objects in question. For example, they might treat priming tests as allowing for the experimental investigation of implicit memory. In doing so, they take advantage of conceptual assumptions about their objects. But what is the ontological status of such objects? Are they cognitive kinds? And if so, what kinds of things are cognitive kinds? In my talk I will argue that cognitive kinds are cognitive-behavioral whole-organism capacities, which are comprised of multiple phenomena, including (but not limited to) behavioral phenomena. With this I depart from the assumption that the behavioral criteria by which cognitive kinds are empirically individuated are mere epistemic vehicles that aid in the investigation of cognitive kinds. Rather, they are part of what it is to be such a kind. While I take cognitive kinds to be sustained by the causal structure of the world, I argue that they are not uniquely determined by neural mechanisms. My account of cognitive kinds is relational in that I claim that cognitive kinds are constituted relative to our sensory-conceptual apparatus and maintained by our causal practices surrounding cognitive kinds. While this analysis is conventionalist, I will argue that this does not imply an antirealism about cognitive kinds.
Andrea Stocco (University of Washington), “Testing and Refining Cognitive Ontologies: From Cognitive Architectures to Large Scale Analysis of the Human Connectome”
Co-Authors: John Laird (University of Michigan), Christian Lebiere (Carnegie Mellon University), Paul Rosenbloom (University of Southern California)
ABSTRACT: Historically, the cognitive neurosciences have drawn from three approaches to define their ontologies: (1) constructs from cognitive psychology, which are usually derived from top-down intuitions about the overall organization of cognition; (2) computational models, which produce well-defined formal models of specific processes; and, more recently (3) large-scale neuroimaging data, which can be used to argue for the identification of large brain architectures from patterns of connectivity. These approaches have yet to converge towards a unified and agreed-upon set of constructs.
Here, we describe an integrative approach that leverages the strengths of all three. First, a consensus model was derived from an analysis of successfully developed cognitive architectures. Focusing on functioning architectures reduces the variability associated with cognitive constructs and provides a preliminary list of “tried and true” mechanisms for human-level cognition. This “Common Model of Cognition” (CMC) is based on a minimal list of five components (perception, action, working memory, declarative memory, and procedural memory) and their interconnections. To partially validate the CMC, we analyzed fMRI data from 200 participants from the Human Connectome Project, using tasks that cover a representative range of cognitive domains (language, mathematics, working memory, relational reasoning, social cognition, emotional inference, and decision-making). The CMC components were identified with functionally homologous brain regions using an iterative procedure that, starting with large-scale, a priori localization assumptions, (e.g., working memory is mapped to fronto-parietal regions), proceeded through the identification of task-specific and individual-specific regions of interest.
Their communication pathways between components were then translated into predicted patterns of effective connectivity. The resulting dynamic model was implemented and fitted using Dynamic Causal Modeling and compared against alternative architectures using a Bayesian approach. We suggest that a successful ontology should explain brain activity equally well across domains (generality) and better than any other model, even specialized ones (comparative superiority). In fact, our results show that, in all cases, the CMC outperforms all other network architectures, both within each domain and across all tasks.
Thus, we conclude that the CMC provides an existence proof of how cognitive, computational, and neuroscience approaches can be integrated and evidence that a minimal cognitive ontology might exist.
Oct 12: Michael Anderson (Western University) and Paul Cisek (University of Montréal), “Two Approaches to Reforming the Taxonomy of Cognitive Neuroscience”
ABSTRACT: In previous work (Anderson 2014; 2015) I have argued that the taxonomy of psychology (at least a neuroscientifically-relevant psychology) is in need of reform. This is for a host of reasons, not the least of which is that the taxonomy of cognitive neuroscience was borrowed wholesale from cognitive psychology, against which I have two main complaints: (1) it is explicitly rooted in a version of the computational theory of mind that is, I believe, demonstrably and deeply flawed and (2) it was formulated as an approach to the study of mind that would be autonomous from and unconstrained by the study of the brain. Both create problems for the project of understanding the neural underpinnings of cognition and behavior.
To pursue the hoped-for reforms, I have been an advocate for a data-driven approach wherein large collections of neuroimaging results, which give us information about the patterns of neural activation observed in different experimental contexts, can be subjected to various forms of mathematical analysis. The motivating idea is that analyzing the observed patterns might reveal similarities and differences in neural activation across different contexts that could indicate heretofore unnoticed similarities and differences in psychological constructs, and potentially lead to a reform of the constructs in question. Importantly, the new constructs would be explicitly constrained by, because partly derived from, neuroscientific investigation.
In the work mentioned above, I also raised the hope that additional insight into the best psychological categories might come from close attention to evolutionary biology.
This is a suggestion I did not myself pursue. However, in recent work Paul Cisek (2019) has taken up this challenge by developing an approach to building a neuroscientifically- grounded psychology he calls “phylogenetic refinement”. The idea here is to closely examine what we know about the phylogentic origins of behavioral control systems, and see how these control systems were developed, exapted, and reused in evolutionary descendants. For instance, the evolutionary novelties that emerged some 600 myo in cephalates included lateralized eyes and neural circuitry allowing for oriented approach and escape behaviors. These circuits are known to have been preserved in modern animals (e.g. mice), presumably exist in many other species, and may be used not just the capture of food and avoidance of predators, but also may have been exapted for more complex social behaviors. (It is not clear if the mechanism of exaptation involves literal reuse of the same circuits or instead the deployment of a homologous structure; this is one issue to be explored). The psychology that results from this sort of analysis would be centered around constructs like approach, escape, explore, exploit, and evaluate for which the (original) neural circuitry is known.
This talk will compare and contrast these two approaches, analyzing the promise and limitations of each.
Vincent Bergeron (University of Ottawa), “Carving the Mind at its Homologous Joints”
ABSTRACT: In this talk, I provide an analysis of the notion of cognitive homology. In contrast with the well-known concept of structural homology in biology—defined as the same structure in different animals regardless of form and function—the proposed notion of cognitive homology captures the idea that the basic cognitive contribution of a given homologous brain structure tends to remain stable over long evolutionary time scales. I then argue that this notion provides a powerful conceptual tool for the study of cognition. Since a cognitive homology will often consist of an evolutionarily conserved relationship between a homologous brain structure and its basic cognitive contribution, such structure-function mappings can be conceived as basic building blocks of animal and human cognition. These basic building blocks, in turn, can be used to construct cognitive ontologies that are well suited to the cognitive neurosciences. To illustrate the usefulness of this approach, I review recent anatomical and functional studies which indicate that the fundamental contribution of Broca’s region to the higher control of action has been conserved throughout primate evolution.
Javier Gomez-Lavin (University of Pennsylvania), “Productive Pessimism and New Ontologies of Cognition”
ABSTRACT: The hope of an easy mapping of psychological function to neural structure has yielded to pessimism in the face of evidence demonstrating that no one region of the brain works in isolation. Focusing on the case study of working memory—our famed ability to keep information in mind—I show that there is no coherent mapping from this psychological construct to a univocal neural structure. The conspicuous lack of progress occasioned by privileging psychological constructs when crafting explanatory ontologies of cognition, forces us to entertain a productive pessimism about their guiding role and the project of mapping their neural realizers. Worries about the smooth mapping of fashionable psychological constructs to neural matter have shadowed the development of neuroscience, but what makes such caution productive? I argue that if we take seriously the possibility of mereological mismatches between agent-level, intuitive descriptions of cognition (e.g., attention, working memory, executive control) and their many, likely overlapping, neural realizers, we can begin to structure ontologies that respect the dynamic, noisy, and multifunctional operation of the brain. To put it in terms of working memory, there are many ways to keep something in mind.
Thursdays 10-11:30 am Eastern Time Zone
Brian Bruya (Eastern Michigan University), “Diverse Origins of Cognitive Ontology”
ABSTRACT: The current method of testing inherited notions of cognitive ontology through the instruments and methods of cognitive neuroscience is vulnerable to confirmation bias and resistant to contrasting alternatives. This talk will discuss effortless attention as an example of challenging a dominant paradigm of cognitive ontology from the perspective of non-standard sources and will discuss some of the findings from very recent attempts to study effortless attention and introduce it as a legitimate element of cognitive ontology. I will first introduce the notion of attention as effort, advocated by Daniel Kahneman (1973), a position that stands as the current paradigm of attention in the cognitive sciences. I will then problematize this position by introducing a notion of effortlessness from classical Daoism, then an allied notion from the psychologist Mihaly Csikszentmihalyi. Finally, I will turn my focus to the concept of postvoluntary attention, first postulated by the Russian psychologist Nikolaj Dobrynin. I will introduce this idea and discuss its relevance to current research in attention and education.
Julia Haas (Rhodes College), “Computational Cognitive Ontology”
ABSTRACT: Much of the recent literature on cognitive ontology proposes a kind of cognitive ontological ‘splitting,’ e.g., when the neural reuse hypothesis proposes that activities of different brain regions recombine to support performance across many different task domains (Anderson 2010, 2014). This paper explores a new way of putting things back together (i.e., ‘lumping’), namely, by taxonomizing computational solutions in the brain rather than relying on traditional psychological capacities. Specifically, I argue that the brain consistently reuses computational solutions to solve certain types of problems, e.g., as when a small suite of reward functions guides selection between competing states of affairs in processes ranging from sensation to practical deliberation. Hence, whereas neural reuse suggests that a single neural structure can support several discrete psychological functions, so computational reuse proposes that seemingly discrete psychological functions are supported by a single computational solution. I then illustrate how we can leverage this notion of computational reuse to revise our extant cognitive ontologies, for example, contrasting the traditional categories of ‘desire’ and ‘motivation’ with the more computationally-informed notion of ‘valuation.’
Nov 5: Jackie Sullivan (Western University), “Cognitive Ontologies, Epistemic Communities and Coordinated Pluralism”
ABSTRACT: Is there only one right cognitive ontology, i.e., the one that carves the mind at the brain’s joints? In this talk, appealing to case studies from areas of cognitive neuroscience that seek to link rodent and human data, I argue that cognitive ontology pluralism is a more realistic proposal. Given that different areas of cognitive neuroscience have different predictive and explanatory goals, experiments in these different areas will be subject to different kinds of epistemic constraints. Moreover, satisfying these constraints requires the formation of what I dub “epistemic communities” whose members collaboratively implement a number of knowledge-building strategies in an effort to achieve their shared predictive or explanatory goals. If this picture is correct, it is suggestive that there will be as many cognitive ontologies as there are epistemic communities in cognitive neuroscience. I contend that this is not a problem just so long as this plurality is used as a corrective to illuminate the limitations of each ontology and thus improve our overall understanding of the mind-brain.
Yoed Kenett (University of Pennsylvania), “Developing a Neurally Informed Ontology of Creativity Measurement,”
Co-Authors: David J. M. Kraemer (Dartmouth College), Katherine L. Alfred (Dartmouth College), Griffin A. Colaizzi (Georgetown University), Robert A. Cortes (Georgetown University), & Adam E. Green (Georgetown University
ABSTRACT: A central challenge for creativity research is to establish a mapping between constructs and measures. A related challenge is the lack of consistency of measures used by different researchers, which hinders progress toward shared understanding of cognitive and neural components of creativity. New resources for aggregating neuroimaging data, and the emergence of methods for identifying structure in multivariate data, present the potential for new approaches to address these challenges. Identifying meta-analytic structure (i.e., similarity) in neural activity associated with creativity tasks might enable identification of a set of tasks that best reflects the similarity among a set of creativity-relevant constructs. Here, we demonstrated initial proof-of-concept for such an approach. We surveyed creativity researchers to build a model of similarity between creativity-relevant constructs. Next, we used NeuroSynth meta-analytic software to generate maps of neural activity robustly associated with tasks intended to measure the same set of creativity-relevant constructs. A representational similarity analysis-based approach revealed that the fit between these models was stronger for some constructs and weaker for others. Critically, we identified particular constructs—and particular tasks measuring those constructs—that positively or negatively impacted the model fit.
Marco Viola (University of Turin), “A Neural-based Assessment of Basic Emotion Theory: Accept, Reject, or Revise and Resubmit?”
ABSTRACT: After decades of disagreement on psychological grounds, the debate over the existence of Basic Emotions has moved to neuroscience. While it is generally agreed that Basic Emotions cannot be mapped on to dedicated neural regions, they do seem to correlate with sets of regions. Some researchers argue that this latter kind of mapping vindicates the existence of Basic Emotions, others (such as psychological constructionists) argue for their elimination from our mental ontology.
In my talk, I propose that this disagreement is due to different attitudes towards how we should approach brain-based reforms of cognitive ontology: researchers friendly to Basic Emotions are ‘conservatives’, i.e. they use brain data to validate psychological categories we already have, while psychological constructionists advocate a ‘radical’ approach, where neural data are seen as a mean to refurnish the cognitive ontology. I discuss some shortcomings of both positions.
With this diagnosis in place, I unpack the macro-question “does the neural evidence support Basic Emotions?” into two smaller questions, namely “is Basic Emotion Theory a viable research program?” and “are the six categories proposed by Ekman vindicated?”. By responding “yes” and “no”, respectively, I endorse a ‘moderate’ approach to revisions of our cognitive ontology of emotions.
Dec 3: Joe McCaffrey (University of Nebraska, Omaha), “Atlas of the Mind: Neural Degeneracy and Pluralistic Ontologies”
ABSTRACT: The recent debate on “Cognitive Ontology” has at its core the assumption that there is a single, correct taxonomy of mental kinds. In this talk, I examine the assumption that there could be a single “map”–instantiated in a database or taxonomy–of mental kinds. I argue that this assumption is mistaken, and illustrate how neural degeneracy (i.e. different brain structures perform “the same” function at different times) in neurotypical individuals and patient populations creates a need for plural taxonomies of mental kinds. Thus, the development of cognitive ontologies is more like constructing an “atlas” or collection of maps rather than representing a unified taxonomy.
Dec 10: Carl Craver (Washington University), “Remembering: Epistemic and Empirical”
ABSTRACT: The effort to unify philosophical and scientific theories of remembering is hampered by the fact that “remember” is used in distinct intellectual contexts to describe altogether different sorts of phenomena. These senses of remembering are designed to serve different theoretical and instrumental objectives. They have apparently opposite commitments. Yet I’ll argue these senses of what remembering is are neither in competition nor in tension with one another; there is no intellectual requirement that the forces molding the contours of the concept in one domain must be responsive to the forces molding the contours of the concept in the other. If we give up the idea that these views—one empirical, describing bio-psychological capacities and their mechanisms; the other epistemic, declaring an achievement, a success, in the effort know the past— must either refer to the same thing (as the reductionist would have it) or be in competition with one another (as elminitavists hold), we might begin to sketch an alternative vision for how these two conceptions are related. The cost of failing to mark this intellectual divide is continued equivocation at the nexus of mind and matter. In fact, the equivocation between the empirical and epistemic is not unique to discussions of remembering but infects a raft of terms at least doubly enlisted in distinct intellectual projects. Believe, explain, know, infer, represent, see, and understand, for example, all have empirical and epistemic senses of the sort described here. Viewed from the standpoint of empirical science and the mechanistic norms of theory development, it seems the intellectual choice we confront in each case is between reducing the epistemic notion to the empirical or, failing such reduction, jettisoning the epistemic construct as pretheoretic folk theory, or philosophy in the worst sense.
Once the difference between these ways of using “remember” is acknowledged, however, it’s clear both that and why epistemic remembering is not even plausibly reductively explained by empirical remembering. The thought that such a reduction is desirable and, correlatively, that the impossibility of reduction is problematic for the epistemic conception, rests on the failure to see that they need not be brought into registration with one another to earn their conceptual keep. These are languages in parallel, and the drive to speak them with one voice only muddles the message about how the mind is situated in the causal structure of things. An adequate language would have to provide the resources for an impossible task: deriving a normatively significant distinction from a reductive base described explicitly so as to fail to mark that very distinction. The problem thus articulated shares key elements with other inference barriers discussed in philosophy, such as the projection of future patterns from the past and, perhaps more aptly, the derivation of what ought to be the case from what is, in fact, the case (Restall and Russell 2010; Pigden 2010).
Adina Roskies, Dartmouth College
Trey Boone, University of Pittsburgh
Mazviita Chirimuuta, University of Pittsburgh
Edouard Machery, University of Pittsburgh
Zina Ward, University of Pittsburgh
The Center for Philosophy of Science
Further inquiries may be addressed to Alex Magee (email@example.com).