I have recently been awarded a five-year European Research Council Advanced Investigator Grant for project CONSCIOUS: Explaining conscious experiences in terms of neural mechanisms. The aim is go beyond merely establishing correlations between patterns of brain activity, towards developing and testing explanations of properties of consciousness in terms of neural mechanisms. The approach can be thought of as ‘computational (neuro-)phenomenology’. One strand of CONSCIOUS will focus on machine learning models of predictive perception. A second will focus on new measures of ‘emergence’ applicable to brain dynamics. Open PhD and postdoc positions will be advertised soon.
Being a beast machine
For many years I’ve been developing a theory of consciousness and self which argues that all conscious experiences – of the world and of the self – are rooted in predictive models in the brain that are geared towards keeping the body alive. We are conscious ‘beast machines’, and consciousness has more to do with being alive than with being intelligent. I’ve also explored other aspects of the philosophy of consciousness, including the ‘real problem’ of consciousness, the relationship between predictive processing and consciousness, and the possibility of ‘islands of awareness’ in (for example) brain organoids.
Seth, A.K. (2021). Being You: A New Science of Consciousness. [coming September 2021]
Causality, emergence, complexity
I have long-standing interests in developing measures of causality, emergence, and complexity, that are applicable in neuroscience. Our measures of causality are based on Granger causality and transfer entropy, and our field-leading software is freely available. More recently (following an early start in 2010) we’ve been working on of how large-scale ‘macroscopic’ properties emerge from their microscopic components, in non-spooky ways that can be quantified. We’ve also been working on empirically-applicable measures of ‘integrated information’ in the context of Giulio Tononi’s integrated information theory of consciousness. This research is with Lionel Barnett, Adam Barrett, Guillaume Corlouer, and Nadine Spychala (Sussex) – in collaboration with Pedro Mediano and Dan Bor (Cambridge), and Fernando Rosas (Imperial College London).
Seth, A.K., Barrett, A.B., and Barnett, L.C. (2015). Granger causality analysis in neuroscience and neuroimaging. Journal of Neuroscience 35(8):3293-3297.
Phenomenological control and perceptual experience
People reliably differ in how ‘suggestible’ they are. We are investigating the role of individual differences in suggestibility – what we call phenomenological control – in many common effects in psychology. We’ve found that phenomena including the rubber hand illusion and ‘mirror synaesthesia’ depend substantially on such differences. In fact, the rubber hand illusion may be nothing more than a suggestion effect. We are now investigating how widely these effects apply, and we’re also developing predictive processing models of phenomenological control in terms of ‘top-down’ influences. This work has methodological implications: response to suggestion can be thought of in terms of ‘demand characteristics’, whereby participants respond, or have experiences, in ways implicitly encouraged by the experimenter. This work is led by Zoltán Dienes and Peter Lush, with Warrick Roseboom, Ryan Scott, Federico Micheli, and others.
Lush, P., Botan, V., Scott, R.B., Seth, A.K., Ward, J., and Dienes Z. (2020). Trait phenomenological control predicts experience of mirror synaesthesia and the rubber hand illusion. Nature Communications. 11, 4853.
A key theme in my research is to study how neural predictions shape, or constitute, perceptual experiences. This involves a range of approaches from new theory connecting perceptual phenomenology to particular kinds of predictions; psychophysics and neuroimaging; augmented reality experiments, and a variety of other studies covering topics from visual perception to metacognition. A current area of focus is on measuring the effects of ‘self efficacy’ (how good we think we are at a task) on metacognition and confidence. Many lab members and colleagues contribute to this work, including Maxine Sherman, Clémence Compain, Federico Micheli, Warrick Roseboom, and Keisuke Suzuki.
Seth, A.K. (2014). A predictive processing theory of sensorimotor contingencies: Explaining the puzzle of perceptual presence and its absence in synaesthesia. Cognitive Neuroscience (target article) 5(2):97-118.
Active inference and machine learning
Active inference is an extension of ‘predictive processing’ models of brain function in which agents minimize prediction errors by actions, as well as by updating predictions. We use computational modelling to examine how active inference can account for biases and inaccuracies in perceptual experience, and we’ve been developing new machine learning algorithms that work by active inference. Currently, we’re working on hybrid schemes in which predictions flow bidirectionally in networks supporting both ‘fast’ and ‘slow’ inference. This work is led by Chris Buckley, with Alexander Tschantz, Beren Millidge, and Tomasz Korbek.
Tschantz, A., Seth, A.K., and Buckley, C.L. (2020). Learning action-oriented models through active inference. PLoS Computational Biology. 16(4):e1007805.
Time perception and memory
We have eyes to see, ears to hear, but how do we perceive time? Instead of relying on an ‘inner clock’, time perception may depend on how the brain processes salient changes in sensory data. Led by Warrick Roseboom, we’ve tested this idea by combining psychophysics and computational modelling, and (more recently) neuroimaging. We are currently extending our approach to incorporate a model of episodic memory. The team also includes Maxine Sherman and Reny Baykova at Sussex, as well as collaborators Zafeirious Fountas (UCL) and Kyriacos Nikiforou and Murray Shanahan (Imperial College London).
Roseboom, W., Fountas, Z., Nikiforou, K., Bhowmik, D., Shanahan, M.P., and Seth, A.K.
(2019). Activity in perceptual classification networks as a basis for human subjective time perception. Nature Communications. 10:267.
Synaesthesia, psychedelics, stroboscopic hallucinations
There is much to be learned by studying unusual conscious states. We’ve pioneered the induction of synaesthesia-like experiences through perceptual training, and we’ve unveiled key changes in neural dynamics in the psychedelic state – including increases in the diversity of neural signals, and decreases in information flow. We are currently studying the visual hallucinations that are induced by powerful stroboscopic stimulation. These projects are led by Lionel Barnett and David Schwartzman at Sussex, with many collaborators – including the Centre for Psychedelic Research at Imperial College London, as well as Daniel Bor and Nicholas Rothen for the synaesthesia work.
Barnett, L.C., Carhart-Harris, R., Muthukumaraswamy, S., and Seth, A.K. (2020). Decreased directed functional connectivity in the psychedelic state. Neuroimage. 209:116462.
Extended reality and perceptual presence
Extended Reality (XR) technologies provide enable us to study conscious perception in immersive environments, and to manipulate experiences of embodiment and selfhood. We’ve been using augmented reality to investigate perceptions of agency, and to simulate psychedelic experiences. Currently we are developing new ‘substitutional reality’ methods to shed new light on the perception of ‘reality’ itself. This work is led by Keisuke Suzuki, with Alberto Mariola, Warrick Roseboom, and David Schwartzman.
Suzuki, K., Roseboom, W., Schwartzman, D.J., and Seth, A.K. (2017). The hallucination machine: A novel method for studying the phenomenology of visual hallucination. Scientific Reports 7(1):15982.
Open science is better science. The entire ecosystem of science can be, and must be, improved – from funding, to research practice, to publication, and to impact. With Jakob Hohwy I edit the open-access journal Neuroscience of Consciousness, which now accepts Registered Report submissions (edited by Zoltán Dienes). Through my leadership positions at Sussex and for CIFAR I promote equity, diversity and inclusion in consciousness research. Wherever possible we make all our publications and data freely available, as are our open-source toolboxes (e.g., for Granger causality analysis).
Seth, A.K. (2019). Consciousness: the last 50 years (and the next). Brain and Neuroscience Advances.