Much of our research revolves around the predictive processing framework—the idea that the brain is a sophisticated hypothesis-tester, continually minimising its own prediction error through both perception and action.
We employ theoretical and philosophical analysis, and our methods include behavioural tasks, psychophysics, EEG, and fMRI.
Research areas include:
- Predictive processing, active inference and the free energy principle
- The neuroscience of consciousness and self
- The philosophy, ethics and psychology of contemplative practices
- Agency, decision-making and rationality
- Psychology, psychiatry and neurology (for example, autism, substance use disorder, eating disorders, borderline personality disorder, functional motor disorders, Parkinson's disease).
For an introduction to the notion to predictive processing, see Jakob Hohwy's review paper, New directions in predictive processing, Mind & Language 35(2): 209-223. doi: 10.1111/mila.12281 [preprint]. See also The Predictive Mind (out in Japanese too).
In the gif above, can you guess which square the participant is controlling? In this experiment, participants had to move a mouse to decide which of eight squares they controlled. Knowing which aspects of the environment we have control over and how this is affected by changing uncertainty is incredibly important to our ability to navigate and understand the world. Here, we manipulated variability and volatility in the sensory signal. Using a novel method, we combine eyetracking (as moment-to-moment hypothesis) and mouse movement (as the prior) to quantify a behavioural proxy for prediction error.
Results show that participants move differently depending on the uncertainty as they dynamically close the action-perception loop. Using our prediction error measure, we could investigate patterns around events and show that changes in hypothesis are preceded by an increase in prediction error and followed by a rapid decrease. We also found that when participants judge they have agency, they minimise prediction error most quickly, regardless of accuracy. Finally, we show that autistic traits do not affect accuracy, but do relate to differences in policy selection, smaller differences in prediction error between levels of uncertainty and a lower amplitude in the hypothesis switch ERPE.
Risk perception and personal responsibility during COVID-19: An experimental study of the role of imperative vs reasoning-based communication for self-isolation attitudes.
van Baal, S. T., Walasek, L., Karanfilovska, D., Cheng, A., & Hohwy, J. (preprint). Risk perception, illusory superiority, and personal responsibility during COVID-19: An experimental study of attitudes to staying home. https://doi.org/10.31234/osf.io/s7jeqvan
Want to go for a run, to the shops, or to a bar during COVID-19? Should you? If you go, can others go too? We explored if imperative messages (StayHome) aid personal responsibility, or if it’s better to reason about situations (StaySafe). Participants were asked to rate if it is alright to go out for 30 everyday scenarios, framed with imperatives (Stay Home, if you can) or reasoning (Consider how many you’d meet). To see if people find loopholes for themselves we also changed perspective (You vs Somebody). Participants were sensitive to the risk levels of scenarios, and imperative framing was most efficient, especially for low risk scenarios. No loophole effects as yet, but some interesting interaction trends to explore in larger samples later. Men and younger people are more willing to go out (no surprises there). And a few more middle-aged individuals seemed strongly risk-seeking.
What is FEP’s epistemic status? It's a priori principle concerning the conditions of possibility for self-organised existence. Don’t look for empirical evidence for it. The process theories under it can be empirically investigated. Why should we believe FEP then? It delivers an account of normativity (or self-supervision) that is, I believe, new/coherent/adequate (but more discussion needed here!). In a lofty slogan, what a system ‘would’ and ‘should’ do is reduced jointly to the ‘is’ of existence. Process theories that conform to FEP are then likely to describe self-supervised systems and processes. That’s important because those are the systems I think we are particularly interested in.
From allostatic agents to counterfactual cognisers: Active inference, biological regulation, and the origins of cognition
What is the function of cognition? According to philosopher Peter Godfrey-Smith, cognition evolved to deal with environmental complexity. In light of growing evidence that plants and bacteria flexibly modulate their activity in response to environmental dynamics, Godfrey-Smith’s thesis would seem to imply that many (if not all) organisms are in fact cognitive systems. Here, we explore an alternative account that seeks to differentiate biological cognition from more pervasive kinds of adaptive plasticity.
Drawing on contemporary advances in theoretical biology and computational neuroscience, we cash these distinctions out in terms of different kinds of generative models, and the representational and uncertainty-resolving capacities they afford. If this analysis is on the right track, cognition is characterised by the ability to detach from one’s present engagements with the world in order to entertain alternative possible realities — a process we refer to as counterfactual active inference.
Raised visual contrast thresholds with intact attention and metacognition in fuctional motor disorder.
Functional motor disorders (FMDs) are a broad subcategory of neurological conditions that affect bodily movement. They continue to puzzle medical science because, unlike well-recognised disorders such as Parkinson’s disease, their origin cannot be traced to distinct areas of the brain or nervous system. Consequently, FMDs are challenging to diagnose and individuals that suffer from these conditions can wait years before they are directed to appropriate treatments. This situation has motivated novel accounts from cognitive science, including the predictive processing framework that theorises FMDs might arise from a subtle but debilitating dysfunction in how the brain perceives the world and acts on these perceptual signals. But, this theory was untested… Until now.
In an interdisciplinary study involving philosophers, psychophysicists, and neurologists, we examined visual perception in three carefully-selected groups: patients with FMDs, patients with traditional organic motor disorders, and healthy control participants matched across several dimensions including age and education. We found that patients with FMDs were identical to control participants across many of our measures. However, this was contingent on correcting a more fundamental difference between the groups—patients required a stronger perceptual signal than control participants. In other words, we had to substantially increase the contrast of the visual target in order for patients with FMDs to perform the task. Our study highlights that perception is an area of interest for understanding the origin of functional motor disorders and that visual perceptual tasks may one-day be an extra tool that neurologists can use to diagnose these conditions and provide individuals that suffer from these conditions with the correct treatment.
The stereotype of autism spectrum conditions (ASC or ‘autism’) focuses on the social and communicative elements of the diagnostic criteria. In this review, we step back from autism as a social and communicative disorder and focus on the autistic self. The autistic self is a key component of the condition which has nevertheless received comparatively little attention. We provide a taxonomy for experimental paradigms in the cognitive sciences that aim to address questions related to the self. We articulate reasons based on domain-general cognitive mechanisms, autobiography and historical conceptions for why the self might differ in ASC. We conclude with elucidating the implications of a predictive processing account of autism on conceptualising the autistic self and how this fits with existing literature, with a focus on context sensitivity, model complexity, learning, integration, active inference and precision. This opens up large scope for future research on unique differences in the autistic self, which could be extended as a framework for understanding the condition as a whole in a new and unified way.
In visual search of natural scenes, differentiation of briefly fixated but task-irrelevant distractor items from incidental memory is often comparable to explicit memorization. However, many characteristics of incidental memory remain unclear, including the capacity for its conscious retrieval. Here, we examined incidental memory for faces in either upright or inverted orientation using Rapid Serial Visual Presentation (RSVP). Subjects were instructed to detect a target face in a sequence of 8-15 faces cropped from natural scene photographs (Experiment 1). If the target face was identified within a brief time window, the subject proceeded to an incidental memory task. Here, subjects used incidental memory to discriminate between a probe face (a distractor in the RSVP stream) and a novel, foil face. In Experiment 2 we reduced scene-related semantic coherency by intermixing faces from multiple scenes and contrasted incidental memory with explicit memory; a condition where subjects actively memorised each face from the sequence without searching for a target. In both experiments, we measured objective performance (Type 1 AUC) and metacognitive accuracy (Type 2 AUC) revealing sustained and consciously accessible incidental memory for upright and inverted faces. In novel analyses of face categories, we examined whether accuracy or metacognitive judgements are affected by shared semantic features (i.e., similarity in gender, race, age). Similarity enhanced the accuracy of incidental memory discriminations but did not influence metacognition. We conclude that incidental memory is sustained and consciously accessible, is not reliant on scene contexts, and not enhanced by explicit memorization.
Neural markers of predictive coding under perceptual uncertainty revealed with hierarchical frequency tagging
Gordon, N., Koenig-Robert, R., Tsuchiya, N., van Boxtel, J. J., & Hohwy, J. (2017). Elife, 6, E22749.
There is a growing understanding that both top-down and bottom-up signals underlie perception. But it is not known how these signals integrate with each other and how this depends on the perceived stimuli’s predictability. ‘Predictive coding’ theories describe this integration in terms of how well top-down predictions fit with bottom-up sensory input. Identifying neural markers for such signal integration is therefore essential for the study of perception and predictive coding theories. To achieve this, we combined EEG methods that preferentially tag different levels in the visual hierarchy. Importantly, we examined intermodulation components as a measure of integration between these signals. Our results link the different signals to core aspects of predictive coding, and suggest that top-down predictions indeed integrate with bottom-up signals in a manner that is modulated by the predictability of the sensory input, providing evidence for predictive coding and opening new avenues to studying such interactions in perception.
Bayne, T., Hohwy, J., Owen, A. (2016). Trends in Cognitive Sciences 20(6): 405-413. The notion of a level of consciousness is a key construct in the science of consciousness. Not only is the term employed to describe the global states of consciousness that are associated with post-comatose disorders, epileptic absence seizures, anaesthesia, and sleep, it plays an increasingly influential role in theoretical and methodological contexts. However, it is far from clear what precisely a level of consciousness is supposed to be. This paper argues that the levels-based framework for conceptualizing global states of consciousness is untenable and develops in its place a multidimensional account of global states.
Ding, C., Palmer, C., Hohwy, J., Youssef, G., Paton, B., Tsuchiya, N., Stout, J., Thyagarajan, D. (2018). Scientific Reports Vol. 8, Article: 13842 (2018).
Parkinson’s disease (PD) alters cortico-basal ganglia-thalamic circuitry and susceptibility to an illusion of bodily awareness, the Rubber Hand Illusion (RHI). Bodily awareness is thought to result from multisensory integration in a predominantly cortical network; the role of subcortical connections is unknown. We studied the effect of modulating cortico-subcortical circuitry on multisensory integration for bodily awareness in 24 PD patients treated with subthalamic nucleus (STN) deep brain stimulation (DBS), in comparison to 21 healthy volunteers, using the RHI experiment. Typically, synchronous visuo-tactile cues induce a false perception of touch on the rubber hand as if it were the subject’s hand, whereas asynchronous visuo-tactile cues do not. However, we found that in the asynchronous condition, patients in the off-stimulation state did not reject the RHI as strongly as healthy controls; patients’ rejection of the RHI strengthened when STN-DBS was switched on, although it remained weaker than that of controls. Patients in the off-stimulation state also misjudged the position of their hand, indicating it to be closer to the rubber hand than controls. However, STN-DBS did not affect proprioceptive judgements or subsequent arm movements altered by the perceptual effects of the illusion. Our findings support the idea that the STN and subcortical connections have a key role in multisensory integration for bodily awareness. Decision-making in multisensory bodily illusions is discussed.