Is there a single principle by which neural operations can account for perception, cognition, act... more Is there a single principle by which neural operations can account for perception, cognition, action, and even consciousness? A strong candidate is now taking shape in the form of "predictive processing". On this theory, brains engage in predictive inference on the causes of sensory inputs by continuous minimization of prediction errors or informational "free energy". Predictive processing can account, supposedly, not only for perception, but also for action and for the essential contribution of the body and environment in structuring sensorimotor interactions. In this paper I draw together some recent developments within predictive processing that involve predictive modelling of internal physiological states (interoceptive inference), and integration with "enactive" and "embodied" approaches to cognitive science (predictive perception of sensorimotor contingencies). The upshot is a development of predictive processing that originates, not in Helmholtzian perception-as-inference, but rather in 20 th-century cybernetic principles that emphasized homeostasis and predictive control. This way of thinking leads to (i) a new view of emotion as active interoceptive inference; (ii) a common predictive framework linking experiences of body ownership, emotion, and exteroceptive perception; (iii) distinct interpretations of active inference as involving disruptive and disambiguatory-not just confirmatory-actions to test perceptual hypotheses; (iv) a neurocognitive operationalization of the "mastery of sensorimotor contingencies" (where sensorimotor contingencies reflect the rules governing sensory changes produced by various actions); and (v) an account of the sense of subjective reality of perceptual contents ("perceptual presence") in terms of the extent to which predictive models encode potential sensorimotor relations (this being "counterfactual richness"). This is rich and varied territory, and surveying its landmarks emphasizes the need for experimental tests of its key contributions.
Too many ghosts in the machine [Review] Beth Singler (2019) Ghost in the Machine Article (Accepte... more Too many ghosts in the machine [Review] Beth Singler (2019) Ghost in the Machine Article (Accepted Version) http://sro.sussex.ac.uk Seth, Anil (2019) Too many ghosts in the machine [Review] Beth Singler (2019) Ghost in the Machine. Nature Machine Intelligence.
Despite rapidly advancing work across the sciences, the means by which neurochemical brain activi... more Despite rapidly advancing work across the sciences, the means by which neurochemical brain activity engenders subjective conscious experience (such as the pain of a stubbed toe, or of a lingering regret) can still seem entirely mysterious, and perhaps requiring scientific revolution rather than evolution. Quantum 'theories' of consciousness capitalize on this suspicion but, in the absence of generating testable predictions or unifying insights, they presently do little to advance our understanding of this fundamental biological property. In their review, Baars and Edelman [2] (this issue) provide an important service by reminding us of the wealth of biological evidence that bears on the problem of consciousness, as well as showcasing the rich repertoire of experimental methodologies supporting further advances. In doing so they call into question the need for a quantum 'revolution' in consciousness science. Their approach reflects what is historically a successful strategy with respect to apparently mysterious phenomena: describe what you are trying to explain, and then come up with testable explanations. This may sound trite, but it is not meant to. Just this strategy characterizes how the biology of life has (successfully) unfolded: What was at one time monolithic and mysterious has become increasingly understood as a constellation of biological processes each accounting for identifiable and partially separable features: metabolism, reproduction, homeostasis, development, and the like [10]. Similarly, as well described by Baars and Edelman, subjective conscious experience has many separable phenomenal properties (such as unity, a first-person-perspective, and tightly-bound multimodal contents; see also [16]) that may each depend on distinct but overlapping neural mechanisms. The challenge for consciousness science is to unravel these mechanisms in ways which explain-or 'account for'-the corresponding phenomenal features [5,7,14,15]. This process can be characterized as the search for so-called 'explanatory correlates' of consciousness (ECCs, [14]), which represents a development of the 'neural correlates of consciousness' (NCC) approach characterizing early work in
As the Simulation of Adaptive Behaviour (SAB) field continues to mature, it is essential that gen... more As the Simulation of Adaptive Behaviour (SAB) field continues to mature, it is essential that general methodological positions become elaborated into practical programmes of research. This paper describes how a particular flavour of SAB modelling-the use of genetic algorithms to design situated agent (animat) architectures-can effectively complement 'optimal foraging theory', as it is understood in theoretical biology. This allows several fundamental problems that arise directly out of the framework of orthodox optimal foraging theory to be addressed, but, as with any trade-off, is not without disadvantages of its own.
I introduce a quantitative measure of autonomy based on a time series analysis adapted from 'Gran... more I introduce a quantitative measure of autonomy based on a time series analysis adapted from 'Granger causality'. A system is considered autonomous if prediction of its future evolution is enhanced by considering its own past states, as compared to predictions based on past states of a set of external variables. The proposed measure, Gautonomy, amplifies the notion of autonomy as 'self-determination'. I illustrate G-autonomy by application to example time series data and to an agent-based model of predator-prey behaviour. Analysis of the predator-prey model shows that evolutionary adaptation can enhance G-autonomy.
Accounts of predictive processing propose that conscious experience is influenced not only by pas... more Accounts of predictive processing propose that conscious experience is influenced not only by passive predictions about the world, but also by predictions encompassing how the world changes in relation to our actions-that is, on predictions about sensorimotor contingencies. We tested whether valid sensorimotor predictions, in particular learned associations between stimuli and actions, shape reports about conscious visual experience. Two experiments used instrumental conditioning to build sensorimotor predictions linking different stimuli with distinct actions. Conditioning was followed by a breaking continuous flash suppression task, measuring the speed of reported breakthrough for different pairings between the stimuli and prepared actions, comparing those congruent and incongruent with the trained sensorimotor predictions. In Experiment 1, counterbalancing of the response actions within the breaking continuous flash suppression task was achieved by repeating the same action within each block but having them differ across the two blocks. Experiment 2 sought to increase the predictive salience of the actions by avoiding the repetition within blocks. In Experiment 1, breakthrough times were numerically shorter for congruent than incongruent pairings, but Bayesian analysis supported the null hypothesis of no influence from the sensorimotor predictions. In Experiment 2, reported conscious perception was significantly faster for congruent than for incongruent pairings. A meta-analytic Bayes factor combining the two experiments confirmed this effect. Altogether, we provide evidence for a key implication of the action-oriented predictive processing approach to conscious perception, namely that sensorimotor predictions shape our conscious experience of the world.
This paper describes a new approach for promoting the evolution of relatively complex behaviours ... more This paper describes a new approach for promoting the evolution of relatively complex behaviours in evolutionary robotics, based on the use of noise in simulation. A`homing navigation' behaviour is evolved (in simulation) for the Khepera mobile robot, and it is shown that high noise levels in the simulation promote the evolution of relatively complex behavioural and neural dynamics. It is also demonstrated that simulation noise can actually accelerate arti cial evolution.
The clinical assessment of non-communicative brain damaged patients is extremely difficult and th... more The clinical assessment of non-communicative brain damaged patients is extremely difficult and there is a need for paraclinical diagnostic markers of the level of consciousness. In the last few years, progress within neuroimaging has led to a growing body of studies investigating vegetative state and minimally conscious state patients, which can be classified in two main approaches. Active neuroimaging paradigms search for a response to command without requiring a motor response. Passive neuroimaging paradigms investigate spontaneous brain activity and brain responses to external stimuli and aim at identifying neural correlates of consciousness. Other passive paradigms eschew neuroimaging in favour of behavioural markers which reliably distinguish conscious and unconscious conditions in healthy controls. In order to furnish accurate diagnostic criteria, a mechanistic explanation of how the brain gives rise to consciousness seems desirable. Mechanistic and theoretical approaches could also ultimately lead to a unification of passive and active paradigms in a coherent diagnostic approach. In this paper, we survey current passive and active paradigms available for diagnosis of residual consciousness in vegetative state and minimally conscious patients. We then review the current main theories of consciousness and see how they can apply in this context. Finally, we discuss some avenues for future research in this domain.
Science and art have long recognised that perceptual experience depends on the involvement of the... more Science and art have long recognised that perceptual experience depends on the involvement of the experiencer. In art history, this idea is captured by Ernst Gombrich's 'beholder's share'. In neuroscience, it traces to Helmholtz's concept of 'perception as inference', which is enjoying renewed prominence in the guise of 'prediction error minimization' or the 'Bayesian brain'. The shared idea is that our perceptual experience-whether of the world, of ourselves, or of an artwork-depends on the active 'top-down' interpretation of sensory input. Perception becomes a generative act, in which perceptual, cognitive, affective, and sociocultural expectations conspire to shape the brain's 'best guess' of the causes of sensory signals. In this paper, I explore the parallels between the Bayesian brain and the beholders' share, illustrated, somewhat informally, with examples from Impressionist, Expressionist, and Cubist art. By connecting phenomenological insights from these traditions with the cognitive neuroscience of predictive perception, I outline a reciprocal relationship in which art reveals phenomenological targets for neurocognitive accounts of subjectivity, while the concepts of predictive perception may in turn help make mechanistic sense of the beholder's share. This is not standard neuroaesthetics-the attempt to discover the brain basis of aesthetic experience-nor is it any kind of neuro-fangled 'theory of art'. It is instead an examination of one way in which art and brain science can be equal partners in revealing deep truths about human experience.
A recent report by Persaud et al. [Persaud, N., McLeod, P. & Cowey, A. (2007). Post-decision wage... more A recent report by Persaud et al. [Persaud, N., McLeod, P. & Cowey, A. (2007). Post-decision wagering objectively measures awareness. Nature Neuroscience 10, 257-261] addresses a fundamental issue in consciousness science: the experimental measurement of conscious content. The authors propose a novel technique, 'post-decision wagering', in which subjects place bets on the correctness of decisions or discriminations. In this note, I critique the authors' claim that their method ''measures awareness directly''.
Granger causality is a statistical concept of causality that is based on prediction. According to... more Granger causality is a statistical concept of causality that is based on prediction. According to Granger causality, if a signal X 1 "Granger-causes" (or "G-causes") a signal X 2 , then past values of X 1 should contain information that helps predict X 2 above and beyond the information contained in past values of X 2 alone. Its mathematical formulation is based on linear regression modeling of stochastic processes (Granger 1969). More complex extensions to nonlinear cases exist, however these extensions are often more difficult to apply in practice. Granger causality (or "G-causality") was developed in 1960s and has been widely used in economics since the 1960s. However it is only within the last few years that applications in neuroscience have become popular.
Consciousness is a key feature of mammalian cognition and revealing its underlying mechanisms is ... more Consciousness is a key feature of mammalian cognition and revealing its underlying mechanisms is one of the most important scientific challenges for the 21st century. In this article I review how computational and theoretical approaches can facilitate a transition from correlation to explanation in consciousness science. I describe progress towards identifying 'explanatory correlates' underlying four fundamental properties characterizing most if not all conscious experiences: (i) the coexistence of segregation and integration in conscious scenes, (ii) the emergence of a subjective first-person perspective, (iii) the presence of affective conscious contents, either transiently (emotion) or as a background (mood) and (iv) experiences of intention and agency that are characteristic of voluntary action. I also discuss how synthetic approaches can shed additional light on possible functions of consciousness, the role of embodiment in consciousness, and the plausibility of constructing a conscious artefact. Keywords Consciousness Á Explanatory correlate Á Causal density Á Complexity Á Perspectivalness Á Emotion Á Volition Á Computational model Á Selfhood Á Emergence Invited article for inaugural issue of Cognitive Computation.
Granger causality is a method for identifying directed functional connectivity based on time seri... more Granger causality is a method for identifying directed functional connectivity based on time series analysis of precedence and predictability. The method has been applied widely in neuroscience, however its application to functional MRI data has been particularly controversial, largely because of the suspicion that Granger causal inferences might be easily confounded by interregional differences in the hemodynamic response function. Here, we show both theoretically and in a range of simulations, that Granger causal inferences are in fact robust to a wide variety of changes in hemodynamic response properties, including notably their time-to-peak. However, when these changes are accompanied by severe downsampling, and/or excessive measurement noise, as is typical for current fMRI data, incorrect inferences can still be drawn. Our results have important implications for the ongoing debate about lag-based analyses of functional connectivity. Our methods, which include detailed spiking neuronal models coupled to biophysically realistic hemodynamic observation models, provide an important 'analysisagnostic' platform for evaluating functional and effective connectivity methods.
Philosophical Transactions of the Royal Society B, Nov 19, 2016
We review a recent shift in conceptions of interoception and its relationship to hierarchical inf... more We review a recent shift in conceptions of interoception and its relationship to hierarchical inference in the brain. The notion of interoceptive inference means that bodily states are regulated by autonomic reflexes that are enslaved by descending predictions from deep generative models of our internal and external milieu. This re-conceptualization illuminates several issues in cognitive and clinical neuroscience with implications for experiences of selfhood and emotion. We first contextualize interoception in terms of active (Bayesian) inference in the brain, highlighting its enactivist (embodied) aspects. We then consider the key role of uncertainty or precision and how this might translate into neuromodulation. We next examine the implications for understanding the functional anatomy of the emotional brain, surveying recent observations on agranular cortex. Finally, we turn to theoretical issues, namely, the role of interoception in shaping a sense of embodied self and feelings. We will draw links between physiological homoeostasis and allostasis, early cybernetic ideas of predictive control and hierarchical generative models in predictive processing. The explanatory scope of interoceptive inference ranges from explanations for autism and depression, through to consciousness. We offer a brief survey of these exciting developments. This article is part of the themed issue 'Interoception beyond homeostasis: affect, cognition and mental health'.
Uploads
Papers by anil seth