Spring 2024
Tuesdays 16:00-17:30
| Date | Seminar | Venue |
|---|---|---|
|
Feb 13 |
Where’s the jam? How listeners resolve semantic ambiguity during speech comprehension. Abstract: The goal of my research is to understand how the brain supports rapid and accurate speech comprehension, and to explain variability in the ease and precision across individuals. I have a particular interest in a common challenge that arises during comprehension – how to interpret a speaker’s intended meaning when there is no one-to-one mapping between spoken words and their meaning (e.g. the semantically ambiguous word “jam” can refer to a fruit preserve to spread on bread, an obstruction, or an informal music session). I will present behavioural and neural data from healthy and brain injured individuals that provide insight into: (1) the time course of meaning computations as semantically ambiguous speech unfolds; (2) which cognitive abilities predict successful comprehension of semantically ambiguous speech; and (3) the contribution of distinct brain networks to speech comprehension made more challenging by the presence of semantic ambiguity (a semantic challenge) in contrast to the challenge arising from words that are difficult to hear (a perceptual challenge). I will show that experiments combining methods from psycholinguistics, psychometrics and cognitive neuroscience can advance theoretical accounts of successful speech comprehension and explain why comprehension sometimes fails. |
Fulton Building, FUL-103 Passcode: 874978 |
|
Feb 27 |
Embodied Sensorimotor Hyperintensionality Abstract: This talk will introduce the notion of hyperintensionality - a concept from formal semantics - for a cognitive science audience. Crudely speaking, hyperintensionality can be thought of as relating to how things are represented, rather than what aspects of the world are represented. I'll explore the relevance to cognitive science of this concept from philosophical logic, explaining how and why it relates to a number of apparently disparate ideas. I'll illustrate this by applying a particular sensorimotor information flow methodology (Kolchinsky & Wolpert's "semantic information" framework) to a classically hyperintensional scenario (an arithmetic task). In the process, I'll touch on (among other things):
|
Fulton Building, FUL-103 Passcode: 234841 |
|
Mar 12 |
The Science of Inner Experience Abstract: Inner experience refers to the contents of consciousness such as inner speech, memories, visual imagery, embodied feelings, etc. Traditionally considered impossible to study because of its private nature, inner experience is emerging as a growing subfield of cognitive science, not least because of the development of new methods for investigating it. Above all, the study of inner experience demands to be approached as a collaborative, interdisciplinary endeavour, drawing in the humanities and social sciences as well as psychology, neuroscience, computer science, philosophy and other traditional cognitive science disciplines. In this talk, I describe some recent interdisciplinary explorations of some aspects of inner experience such as inner speech, memory and hallucinations, and how they have found applications in gaming, AI and the creative industries. Charles Fernyhough is Professor of Psychology at Durham University and Director of its Centre for Research into Inner Experience. |
online Passcode: 740396 |
|
Mar 19 |
Exploring What Feels Real: Cognitive and Computational Approaches to Presence Abstract:Recent advances in XR and generative AI technologies have unlocked the potential to create hyper-realistic environments, visuals, and texts, challenging our perception of what is real. This development invites a re-evaluation of mental disorders—such as hallucinations, delusions, and depersonalization—as disorders of 'reality.' This talk outlines cognitive neuroscience and computational approaches to our experience of reality using VR and AI. I begin with the Hallucination Machine, a novel XR platform to simulate key characteristics of psychedelic visual hallucinations, using deep neural networks. Furthermore, we investigate the perceptual veridicality of visual hallucinations across different etiologies and separately synthesize the different levels of veridicality using alternative generative models. We also explore perceptual presence, the sense that an observer and objects exist in the same space, by examining a subject's sensorimotor interactions with their environment. Through manipulating sensorimotor contingencies, we have validated how sensorimotor contingency could influence the visual processing of 3D objects, potentially mediated by perceptual presence. The emergence of XR and AI is not just advancing research tools to investigate conscious experiences, but also dynamically transforming our epistemology of how we understand reality. This highlights the need for interdisciplinary work to uncover how we experience reality. |
Fulton Building - FUL-103 Passcode: 538040 |
|
Mar 26 |
The relevance of communication theory for theories of representation Abstract:Prominent views about representation share an assumption: that mathematical communication theory is blind to representational content. Here I challenge that assumption by rejecting two claims often enlisted in support of it: that Claude Shannon said that the meanings of signals are irrelevant for communication theory (he didn't and they aren't), and that since correlational measures can't distinguish representations from natural signs, communication theory can't distinguish them either (the premise is true but the conclusion is false; no valid argument can link them). I then offer a replacement view, on which teleosemantics endorses communication theorists' attributions of representational content to signals. Finally I discuss consequences for accounts of representation in cognitive science. |
Zoom Passcode: 525471 |
|
Apr 23 POSTPONED |
Scaling Up Consciousness: From Individual to Social Cognition via Active Inference Abstract:This talk explores how the free energy principle and active inference can bridge the gap between individual and social cognition by tracing the way the temporal structure of consciousness scales up from individual agents to social interaction and shared meaning. Building on the inner screen model of consciousness and Husserl's phenomenology of time consciousness, I show how the components of individual experience (primal impression, retention, and protention) map onto the temporal dynamics of an agent's generative model in active inference. I then extend this framework to social cognition, showing how interactions between agents with similar embodiments lead to the emergence of shared protentions and representations. Using category theory and sheaf theory, I formalize how agents' individual generative models can be integrated into a coherent, shared understanding of the world. Finally, I present a variational approach to social scripts, explaining how shared priors and preferences enable coordinated action and become reified into the fabric of social reality itself. This talk aims to demonstrate how the active inference framework, informed by Husserlian phenomenology and category theory, can illuminate the deep continuities between consciousness, cognition, and culture, showing how the fundamental structures of conscious experience scale up to generate the shared world of social meaning and interaction. |
Zoom Passcode: 592877 |
|
Apr 30 |
Does misinformation have fingerprints? Abstract: An influential view in social science claims that misinformation can be identified by detectable "fingerprints," such as emotional, conspiratorial, and polarising language. These alleged fingerprints underlie widely adopted interventions by tech companies, governments, and international organisations aimed at reducing the spread of misinformation and teaching people to identify it. In this talk, I argue that the existence of these fingerprints lacks theoretical and empirical support. First, I argue that misinformation is unlikely to have diagnostic fingerprints because propagandists deliberately mimic reliable communication, and their tactics are highly context-sensitive. Second, I argue that empirical research claiming to support the existence of these fingerprints relies on unrepresentative samples of misinformation, which cannot justify broad generalisations. At best, research and interventions that assume that misinformation has identifiable fingerprints therefore lack a scientific basis. At worst, they lend spurious scientific legitimacy to the personal preferences of researchers; they cast suspicion on legitimate contributions to public debate; and they risk exacerbating the problems they aim to address. |
Fulton Building - FUL-103 Passcode: 036181 |
|
May 7 |
Mr Market on the Couch; Introducing the Market Mind Hypothesis, the Final Frontier of Behavioural Finance Abstract:This presentation introduces the Market Mind Hypothesis (MMH). MMH pushes the respective boundaries of both the extended/distributed mind theory and behavioural finance. Popularly speaking, MMH recognises that Mr Market has a mind, warts and all. In simple terms, MMH formalises what investors have always casually referred to as "the market's mind", including its relationship to the real (physical) economy which showed a dangerous tail-wagging-the-dog dynamic during the Global Financial Crisis (GFC). More technically, MMH states that the market, embodying interacting conscious investors and their technologies, not only distributes their knowledge but also intersubjectively extends their minds, thereby manifesting collective consciousness. Prices form the signatures of this, while market mood is its phenomenal experience. Unfortunately, Mr Market has not been healthy for a while and neither has the wider economic community of which he is part (exemplified by excessive debt, negative yields, and flash crashes). The key cause is, simultaneously, also mainstream economics' "true problem": not recognising, let alone addressing economics' "hard problem". Specifically, driven by physics envy its mechanical "equilibrium" paradigm views and treats (e.g. via monetary policies) the economy as a machine, the market as an automaton, and humans as robots. In short, mainstream economics has made a flawed ontological commitment that is becoming very expensive. This is particularly relevant in light of the almost uncritical embracement of AI and other automation, but also in the context of ESG investing. The presentation will cover this, as well as numerous other topics, including historic origins, Lehman's Lesson, price discovery, and the projects (to be) hosted in our Market Mind Centre to be located within the Edinburgh Futures Institute. |
Jubilee 144 Passcode: 882843 |
|
May 14 |
A Mark of the Non-Cognitive Abstract:In the so-called cognition wars, proponents of the idea of basal-cognition (Lyon et al. 2021) clash with those seeking to restrict the term “cognitive” to systems showing human-like capacities for manipulation of representations (Adams 2018). The impasse over the definition of cognition is unlikely to end soon. Yet, given that some proponents of the expansive definition have sought to ascribe cognition to all living things, and now even to molecular systems beyond biology (McGivern 2020), the question of what counts as non-cognitive becomes pressing. Do the arguments for basal-cognition inevitably devolve into claims for a kind of panpsychism, or is there a principled standard for what counts as indisputably non-cognitive that all parties of the dispute will adhere to? The purpose of this paper is to offer a mark of the non-cognitive – to characterise what it is about certain entities and processes that makes them paradigmatically non-cognitive. The characterisation is not a novel proposal but a clarification of a rare, though implicit, point of consensus amongst scientists researching on non-neural organisms that have been put forward as candidates for basal cognition. I argue that opinion converges on the notion of non-cognitive processes being those that appear to be the result only of proximal, efficient physical-chemical causes. A system whose activity profile is of this sort will be a simple mechanism and will have the passive and inflexible character that is paradigmatically non-cognitive. I go on to discuss how this characterisation of the non-cognitive relates in interesting ways to notions of mechanisms. |
Fulton Building - FUL-113 Passcode: 872756 |
|
June 4 |
Epistemic Emotions Abstract:Many emotions have an epistemic dimension. For example, anxiety implies uncertainty. There are, in addition, emotions that are epistemic in a more fundamental way. These include, doubt, certainty, interest, among others. This presentation presents a definition of epistemic emotions, a taxonomy demonstrating that this is a rich and highly diverse category, a demonstration that the items in this category are genuine emotions, and an exploration of how epistemic emotions may contribute to knowledge. An argument will also be offered for the claim that epistemic emotions are essential for knowledge. At the end, implications for the epistemic status of large language models will be drawn as well. |
Pevensey 1 A6 Passcode: 308004 |
Contact COGS
For suggestions for speakers, contact Simon Bowes
For publicity and questions regarding the website, contact Simon Bowes.
Please mention COGS and COGS seminars to all potentially interested newcomers to the university.
A good way to keep informed about COGS Seminars is to be a member of COGS. Any member of the university may join COGS and the COGS mailing list. Please contact Simon Bowes if you would like ot be added.
Follow us on Twitter: @SussexCOGS