Our third project workshop, entitled Wilding the Predictive Brain: Prediction, Culture and Context, was held on November 1, 2019 at the University of Sussex.
The picture of the human mind as an embodied prediction machine is now the dominant systems-level model in cognitive neuroscience. But neural prediction never happens in a vacuum. Our predictive brains are themselves embedded in layer upon layer of bodily and worldly structure and opportunity. This year the X-SPECT workshop explores these horizons, highlighting the many ways that predictive minds interface and evolve with society, culture and technology.
PROGRAMME
9.00-10.00 Registration Opens, Coffee and Pastries
10.00 - 11.00 Matthew Sims (University of Edinburgh)
11.00 – 12.00 Sarah Garfinkel (University of Sussex)
12.00 - 1.00 Lunch
1.00 – 2.00 Andreas Roepstroff (Aarhus University)
2.00 – 3.00 Katerina Fotopoulou (University College London)
3.00-4.00 Coffee Break & Poster Session
4.00 - 5.00 Jelle Bruineberg (University of Amsterdam)
5.00 – 6.00 Roslyn Moran (University of Bristol)
6.00 - 7.00 Drinks on Campus
7.00 Dinner for Speakers, Chairs, and Organizers.
TALK TITLES, ABSTRACTS & AUDIO RECORDINGS
Matthew Sims (University of Edinburgh)
The Symbiotic Mind: a Pluralist Account of Bio-Cognitive Individuality
Symbiosis, i.e., mutually beneficial heterospecific associations, has been an important tool for thinking about the problem of biological individuality. This problem is that of determining “what, in the living world, constitutes a relatively well delineated and cohesive unit?” (Pradeu, 2016). One account of biological individuality that has emerged in the debate that considerations of symbiosis have contributed to is that of the physiological individual: a biological system made of heterogeneous parts that through various regulatory processes (e.g., metabolism or immune system regulation) maintains itself as a functional whole. Can symbiosis also help to improve our understanding of what constitutes a well-delineated unit of cognition? Cognitive frameworks such as predictive processing and the free energy principle (FEP), often ascribe cognition to a single individual that authors its own sensory states via adaptive behaviour (i.e., active inference), allowing it to maintain its functional and structural integrity. Recently, using the notion of Markov blanket of Markov blankets, Ramstead, et al. (2019) have put forth a multiscale integrationalist view of FEP. According to this view, the boundaries of cognitive systems are neither fixed nor privileged. Although this multiscale integrationalist view allows for a pluralism about where cognitive capacities are realized (e.g., a spider’s web extends the spiders perception-action capacities), it stops short however of telling us anything about what unit to ascribe cognition to (e.g., is there anything beyond the spider that qualifies as a cognitive individual?). Is there any reason to be pluralist about bio-cognitive individuals? Using the nested Markov blanket formalism and assuming that a form of minimal cognition falls out of FEP, I will look at the symbiosis of Vibrio fischeri bacteria and the bobtailed squid as a case study in support of a pluralist account of bio-cognitive individuality. The squid-Vibrio association constitutes a heterogenous functional unit, where the squid and Vibrio reciprocally contribute to the manner in which the other organism engages in active inference, respectively extending cognition both outwards (for the sheltered vibrio) and inwards (for the camouflaged squid), enacting a transient squid-Vibrio generative model. The emergence the symbiotic bio-cognitive individual allows for a minimization of variational free energy that could not otherwise arise at the scale of the individual Vibrio or squid. I shall argue that the multiscale reciprocal active inference that is instantiated in this kind of mutualistic association reveals a conception of the bio-cognitive individual that maps on to a special class of physiological individuals. Since ascribing cognition to such a physiological individual that arises at the level of the symbiotic consortium is no reason to deny cognition to each constituent physiological individual when coupled in reciprocal active inference, a pluralistic account should be favoured.
Katerina Fotopoulou (University College London)
Mentalizing Adiposity: The Development of the Bodily Self and its Disruption in Anorexia Nervosa
In this interdisciplinary talk, I will put forward the idea that eating disorders are best conceptualised as disorders of bodily self-regulation and more specifically the ability to regulate one’s adiposity in the service of long-term and conflictual energy risks and goals. Self concepts (e.g. self-esteem, self-efficacy, self-objectification) have long being associated with eating disorders but there is hardly any understanding about how such higher-order self concepts may relate to the neurobiology of disordered eating. In this talk I will ground these concepts to more fundamental mechanisms of embodied selfhood and mentalisation and I will present developmental and adult, behavioural and neurobiological data on an individual’s sense of their own body relies on how they built active, inferences about the physiological state of their body, which in turns relies on early ‘affectively contingent’ social experiences of feeding and caregiving. I will describe how deficits in this process of embodied mentalisation in disorders such as restrictive Anorexia Nervosa results in a profound, ‘intolerable uncertainty’ about being able to allostatically regulate the body’s adiposity and other homeostatic needs, which in turn results in rigid behavioural eating restrictions as the only active way to feel a sense of ‘control’ and ‘efficacy’ over one’s own body.
Jelle Bruineberg (University of Amsterdam)
Active inference and technological mediation
The initial optimism about the transformative prospects of Internet and digital technology is making way for a critical debate about their impact on our mental lives and society at large. Part of the debate zooms in on the “attention crisis”: the idea that the omnipresence of digital technologies limits our capability to pay attention to those things that matter to us. This debate is premised on a number of questionable assumptions about the nature of attention, agency and the mind-world relationship. In this talk, I will use active inference as well recent work on the technological mediation of intentionality to understand how digital technologies can impact the agency of its user’s.
Roslyn Moran (along with Jonathan Monney, Berk Merza, Maell Cullen; University of Bristol)
Active Inference and Computer Vision: What do the eyes see?
In this talk I will outline the theory of Active Inference and describe how the expected variational free energy over future states can be used by computer vision systems to decide ‘where to look’. I will provide data from human eye movement studies that demonstrates how human vision systems are top-down machines, where movements are deliberative, sparse and goal driven. The theory of Active Inference proposes that all biological agents retain self-ness by minimizing their long-term average surprisal. In information theoretic terms, the Free Energy provides a soluble approximation to this long-term surprise and necessitates the development of a generative model of the environment within the agent itself. The minimization of this quantity via a gradient flow is purported to be the purpose of neuronal activity in the brain and thus provides a mapping from brain activity to their first-principle computations. In order to make decisions the free energy of future expected states must be minimized. In this talk I will develop how this framework can be applied to perform eye-movements. I will show, using the classic Machine Vision dataset – MNIST that humans decide where to look, sample data and draw conclusions in a way commensurate with this theory. I will contrast this scheme to popular machine vision systems that use solely feedforward networks and discuss the implications for human-like computing.
Andreas Roepstroff (Aarhus University)
Experimenting, Experiencing Reflecting: sharing embodied cognition, predicting other minds
How do we get to experience that another person may experience the same world differently?.
I describe a collaboration between artists and researchers aimed at creating novel settings for exchanging embodied cognition. Art installations in a museum setting may offer a privileged space to engage in perspective taking and sharing, and for experiencing and exploring predictive models of other minds.
Sarah Garfinkel (University of Sussex)
Clinical Neuroscience and the Heart-Brain Axis
The picture of the human mind as an embodied prediction machine is now the dominant systems-level model in cognitive neuroscience. But neural prediction never happens in a vacuum. Our predictive brains are themselves embedded in layer upon layer of bodily and worldly structure and opportunity. This year the X-SPECT workshop explores these horizons, highlighting the many ways that predictive minds interface and evolve with society, culture and technology.
PROGRAMME
9.00-10.00 Registration Opens, Coffee and Pastries
10.00 - 11.00 Matthew Sims (University of Edinburgh)
11.00 – 12.00 Sarah Garfinkel (University of Sussex)
12.00 - 1.00 Lunch
1.00 – 2.00 Andreas Roepstroff (Aarhus University)
2.00 – 3.00 Katerina Fotopoulou (University College London)
3.00-4.00 Coffee Break & Poster Session
4.00 - 5.00 Jelle Bruineberg (University of Amsterdam)
5.00 – 6.00 Roslyn Moran (University of Bristol)
6.00 - 7.00 Drinks on Campus
7.00 Dinner for Speakers, Chairs, and Organizers.
TALK TITLES, ABSTRACTS & AUDIO RECORDINGS
Matthew Sims (University of Edinburgh)
The Symbiotic Mind: a Pluralist Account of Bio-Cognitive Individuality
Symbiosis, i.e., mutually beneficial heterospecific associations, has been an important tool for thinking about the problem of biological individuality. This problem is that of determining “what, in the living world, constitutes a relatively well delineated and cohesive unit?” (Pradeu, 2016). One account of biological individuality that has emerged in the debate that considerations of symbiosis have contributed to is that of the physiological individual: a biological system made of heterogeneous parts that through various regulatory processes (e.g., metabolism or immune system regulation) maintains itself as a functional whole. Can symbiosis also help to improve our understanding of what constitutes a well-delineated unit of cognition? Cognitive frameworks such as predictive processing and the free energy principle (FEP), often ascribe cognition to a single individual that authors its own sensory states via adaptive behaviour (i.e., active inference), allowing it to maintain its functional and structural integrity. Recently, using the notion of Markov blanket of Markov blankets, Ramstead, et al. (2019) have put forth a multiscale integrationalist view of FEP. According to this view, the boundaries of cognitive systems are neither fixed nor privileged. Although this multiscale integrationalist view allows for a pluralism about where cognitive capacities are realized (e.g., a spider’s web extends the spiders perception-action capacities), it stops short however of telling us anything about what unit to ascribe cognition to (e.g., is there anything beyond the spider that qualifies as a cognitive individual?). Is there any reason to be pluralist about bio-cognitive individuals? Using the nested Markov blanket formalism and assuming that a form of minimal cognition falls out of FEP, I will look at the symbiosis of Vibrio fischeri bacteria and the bobtailed squid as a case study in support of a pluralist account of bio-cognitive individuality. The squid-Vibrio association constitutes a heterogenous functional unit, where the squid and Vibrio reciprocally contribute to the manner in which the other organism engages in active inference, respectively extending cognition both outwards (for the sheltered vibrio) and inwards (for the camouflaged squid), enacting a transient squid-Vibrio generative model. The emergence the symbiotic bio-cognitive individual allows for a minimization of variational free energy that could not otherwise arise at the scale of the individual Vibrio or squid. I shall argue that the multiscale reciprocal active inference that is instantiated in this kind of mutualistic association reveals a conception of the bio-cognitive individual that maps on to a special class of physiological individuals. Since ascribing cognition to such a physiological individual that arises at the level of the symbiotic consortium is no reason to deny cognition to each constituent physiological individual when coupled in reciprocal active inference, a pluralistic account should be favoured.
Katerina Fotopoulou (University College London)
Mentalizing Adiposity: The Development of the Bodily Self and its Disruption in Anorexia Nervosa
In this interdisciplinary talk, I will put forward the idea that eating disorders are best conceptualised as disorders of bodily self-regulation and more specifically the ability to regulate one’s adiposity in the service of long-term and conflictual energy risks and goals. Self concepts (e.g. self-esteem, self-efficacy, self-objectification) have long being associated with eating disorders but there is hardly any understanding about how such higher-order self concepts may relate to the neurobiology of disordered eating. In this talk I will ground these concepts to more fundamental mechanisms of embodied selfhood and mentalisation and I will present developmental and adult, behavioural and neurobiological data on an individual’s sense of their own body relies on how they built active, inferences about the physiological state of their body, which in turns relies on early ‘affectively contingent’ social experiences of feeding and caregiving. I will describe how deficits in this process of embodied mentalisation in disorders such as restrictive Anorexia Nervosa results in a profound, ‘intolerable uncertainty’ about being able to allostatically regulate the body’s adiposity and other homeostatic needs, which in turn results in rigid behavioural eating restrictions as the only active way to feel a sense of ‘control’ and ‘efficacy’ over one’s own body.
Jelle Bruineberg (University of Amsterdam)
Active inference and technological mediation
The initial optimism about the transformative prospects of Internet and digital technology is making way for a critical debate about their impact on our mental lives and society at large. Part of the debate zooms in on the “attention crisis”: the idea that the omnipresence of digital technologies limits our capability to pay attention to those things that matter to us. This debate is premised on a number of questionable assumptions about the nature of attention, agency and the mind-world relationship. In this talk, I will use active inference as well recent work on the technological mediation of intentionality to understand how digital technologies can impact the agency of its user’s.
Roslyn Moran (along with Jonathan Monney, Berk Merza, Maell Cullen; University of Bristol)
Active Inference and Computer Vision: What do the eyes see?
In this talk I will outline the theory of Active Inference and describe how the expected variational free energy over future states can be used by computer vision systems to decide ‘where to look’. I will provide data from human eye movement studies that demonstrates how human vision systems are top-down machines, where movements are deliberative, sparse and goal driven. The theory of Active Inference proposes that all biological agents retain self-ness by minimizing their long-term average surprisal. In information theoretic terms, the Free Energy provides a soluble approximation to this long-term surprise and necessitates the development of a generative model of the environment within the agent itself. The minimization of this quantity via a gradient flow is purported to be the purpose of neuronal activity in the brain and thus provides a mapping from brain activity to their first-principle computations. In order to make decisions the free energy of future expected states must be minimized. In this talk I will develop how this framework can be applied to perform eye-movements. I will show, using the classic Machine Vision dataset – MNIST that humans decide where to look, sample data and draw conclusions in a way commensurate with this theory. I will contrast this scheme to popular machine vision systems that use solely feedforward networks and discuss the implications for human-like computing.
Andreas Roepstroff (Aarhus University)
Experimenting, Experiencing Reflecting: sharing embodied cognition, predicting other minds
How do we get to experience that another person may experience the same world differently?.
I describe a collaboration between artists and researchers aimed at creating novel settings for exchanging embodied cognition. Art installations in a museum setting may offer a privileged space to engage in perspective taking and sharing, and for experiencing and exploring predictive models of other minds.
Sarah Garfinkel (University of Sussex)
Clinical Neuroscience and the Heart-Brain Axis