Publications

AXIOM: Learning to Play Games in Minutes with Expanding Object-Centric Models

Current deep reinforcement learning (DRL) approaches achieve state-of-the-art performance in various domains, but struggle with data efficiency compared to human learning, which leverages core priors about objects and their interactions. Active inference offers a principled framework for integrating sensory information with prior knowledge to learn a world model and quantify the uncertainty of its own beliefs and predictions. However, active inference models are usually crafted for a single task with bespoke knowledge, so they lack the domain flexibility typical of DRL approaches. To bridge this gap, we propose a novel architecture that integrates a minimal yet expressive set of core priors about object-centric dynamics and interactions to accelerate learning in low-data regimes. The resulting approach, which we call AXIOM, combines the usual data efficiency and interpretability of Bayesian approaches with the across-task generalization usually associated with DRL. AXIOM represents scenes as compositions of objects, whose dynamics are modeled as piecewise linear trajectories that capture sparse object-object interactions. The structure of the generative model is expanded online by growing and learning mixture models from single events and periodically refined through Bayesian model reduction to induce generalization. AXIOM masters various games within only 10,000 interaction steps, with both a small number of parameters compared to DRL, and without the computational expense of gradient-based optimization.

Imaging Brain Structure and Function in Schizophrenia

Attempts to define a personality profile of individuals disposed principally to self-serving and antisocial behaviour can be traced at least as far back as Pinel's concept of 'manie sans delire'.1 Rush2 relabelled it as 'moral alienation of the mind', and contributed strongly to the Anglo-American conception of psychopathy as predominantly antisocial. Prichard3 concerned himself with the question of antisocial actions under the control of the will, and stimulated a debate that shaped the concept of criminal responsibility, though his term 'moral insanity' had a limited lifespan. Subsequently, Partridge4 suggested the designation 'sociopathy' for personality disturbances whose outstanding feature is antisocial behaviour. Cleckley,5 in his monograph The Mask of Sanity, proposed the use of the term 'psychopathy' (which had been coined by Koch,6 in 1891) for individuals showing lifelong maladaptive behaviour not caused by neurosis, psychosis or mental handicap. Cleckley's definitions were developed in the diagnostic and statistical manual of mental disorders (DSM) and International Classification of Diseases (ICD) schemes.

Gamma Oscillations and Neural Field DCMs Can Reveal Cortical Excitability and Microstructure

This paper shows how gamma oscillations can be combined with neural population models and <em>dynamic causal modeling</em> (DCM) to distinguish among alternative hypotheses regarding cortical excitability and microstructure. This approach exploits inter-subject variability and trial-specific effects associated with modulations in the peak frequency of gamma oscillations. Neural field models are used to evaluate model evidence and obtain parameter estimates using invasive and non-invasive gamma recordings. Our overview comprises two parts: in the first part, we use neural fields to <em>simulate</em> neural activity and distinguish the effects of post synaptic filtering on predicted responses in terms of synaptic rate constants that correspond to different timescales and distinct neurotransmitters. We focus on model predictions of conductance and convolution based field models and show that these can yield spectral responses that are sensitive to biophysical properties of local cortical circuits like synaptic kinetics and filtering; we also consider two different mechanisms for this filtering: a nonlinear mechanism involving specific conductances and a linear convolution of afferent firing rates producing post synaptic potentials. In the second part of this paper, we use neural fields <em>quantitatively</em>—to <em>fit</em> empirical data recorded during visual stimulation. We present two studies of spectral responses obtained from the visual cortex during visual perception experiments: in the first study, MEG data were acquired during a task designed to show how activity in the gamma band is related to visual perception, while in the second study, we exploited high density electrocorticographic (ECoG) data to study the effect of varying stimulus contrast on cortical excitability and gamma peak frequency.

Brief Mindfulness Meditation Induces Gray Matter Changes in a Brain Hub

Previous studies suggest that the practice of long-term (months to years) mindfulness meditation induces structural plasticity in gray matter. However, it remains unknown whether short-term (<30 days) mindfulness meditation in novices could induce similar structural changes. Our previous randomized controlled trials (RCTs) identified white matter changes surrounding the anterior cingulate cortex (ACC) and the posterior cingulate cortex (PCC) within 2 to 4 weeks, following 5-10 h of mindfulness training. Furthermore, these changes were correlated with emotional states in healthy adults. The PCC is a key hub in the functional anatomy implicated in meditation and other perspectival processes. In this longitudinal study using a randomized design, we therefore examined the effect of a 10 h of mindfulness training, the Integrative Body-Mind Training (IBMT) on gray matter volume of the PCC compared to an active control-relaxation training (RT). We found that brief IBMT increased ventral PCC volume and that baseline temperamental trait-an index of individual differences was associated with a reduction in training-induced gray matter increases. Our findings indicate that brief mindfulness meditation induces gray matter plasticity, suggesting that structural changes in ventral PCC-a key hub associated with self-awareness, emotion, cognition, and aging-may have important implications for protecting against mood-related disorders and aging-related cognitive declines.

Active Listening

Abstract This paper introduces active listening, as a unified framework for synthesising and recognising speech. The notion of active listening inherits from active inference, which considers perception and action under one universal imperative: to maximise the evidence for our (generative) models of the world. First, we describe a generative model of spoken words that simulates (i) how discrete lexical, prosodic, and speaker attributes give rise to continuous acoustic signals; and conversely (ii) how continuous acoustic signals are recognised as words. The ‘active’ aspect involves (covertly) segmenting spoken sentences and borrows ideas from active vision. It casts speech segmentation as the selection of internal actions, corresponding to the placement of word boundaries. Practically, word boundaries are selected that maximise the evidence for an internal model of how individual words are generated. We establish face validity by simulating speech recognition and showing how the inferred content of a sentence depends on prior beliefs and background noise. Finally, we consider predictive validity by associating neuronal or physiological responses, such as the mismatch negativity and P300, with belief updating under active listening, which is greatest in the absence of accurate prior beliefs about what will be heard next.

I overthink—therefore I am not: Altered Sense of Self and Agency in Depersonalisation Disorder

This paper considers the phenomenology of depersonalisation disorder, in relation to predictive processing and its associated pathophysiology. To do this, we first establish a few mechanistic tenets of predictive processing that are necessary to talk about phenomenal transparency, mental action, and self as subject. We briefly review the important role of ‘predicting precision’ and how this affords mental action and the loss of phenomenal transparency. We then turn to sensory attenuation and the phenomenal consequences of (pathophysiological) failures to attenuate or modulate sensory precision. We then consider this failure in the context of depersonalisation disorder. The key idea here is that depersonalisation disorder reflects the remarkable capacity to explain perceptual engagement with the world via the hypothesis that “I am an embodied perceiver, but I am not in control of my perception”. We suggest that individuals with depersonalisation may believe that ‘another agent’ is controlling their thoughts, perceptions or actions, while maintaining full insight that the ‘another agent’ is ‘me’ (the self). Finally, we rehearse the predictions of this formal analysis, with a special focus on the psychophysical and physiological abnormalities that may underwrite the phenomenology of depersonalisation.

PsPM-SCRV9: Skin conductance responses to visual targets

This dataset includes skin conductance response (SCR) measurements for each of 22 healthy unmedicated participants (11 males and 11 females aged 24.7+/-4.5 years) participating in a visual detection task. Participants are requested to react by a keypress to a target visual stimulus (a red cross) embedded in a stream of distractors (white numbers). Stimuli are presented during 200ms and are spaced by a 800ms blank interval. A baseline period with distractors but no targets concludes the experiment. (This is in contrast to the methods description in Bach et al. (2010), according to which the baseline period was randomly either in the beginning or at the end of the experiment. This discrepancy was caused by an error in the code that controlled the experiment presentation.)

The hierarchically mechanistic mind: an evolutionary systems theory of the human brain, cognition, and behavior

The purpose of this review was to integrate leading paradigms in psychology and neuroscience with a theory of the embodied, situated human brain, called the Hierarchically Mechanistic Mind (HMM). The HMM describes the brain as a complex adaptive system that functions to minimize the entropy of our sensory and physical states via action-perception cycles generated by hierarchical neural dynamics. First, we review the extant literature on the hierarchical structure of the brain. Next, we derive the HMM from a broader evolutionary systems theory that explains neural structure and function in terms of dynamic interactions across four nested levels of biological causation (i.e., adaptation, phylogeny, ontogeny, and mechanism). We then describe how the HMM aligns with a global brain theory in neuroscience called the free-energy principle, leveraging this theory to mathematically formulate neural dynamics across hierarchical spatiotemporal scales. We conclude by exploring the implications of the HMM for psychological inquiry.

Statistical parametric mapping (SPM) of mucociliary clearance of inhaled radioaerosols.

Federated inference and belief sharing

This paper concerns the distributed intelligence or federated inference that emerges under belief-sharing among agents who share a common world-and world model. Imagine, for example, several animals keeping a lookout for predators. Their collective surveillance rests upon being able to communicate their beliefs-about what they see-among themselves. But, how is this possible? Here, we show how all the necessary components arise from minimising free energy. We use numerical studies to simulate the generation, acquisition and emergence of language in synthetic agents. Specifically, we consider inference, learning and selection as minimising the variational free energy of posterior (i.e., Bayesian) beliefs about the states, parameters and structure of generative models, respectively. The common theme-that attends these optimisation processes-is the selection of actions that minimise expected free energy, leading to active inference, learning and model selection (a.k.a., structure learning). We first illustrate the role of communication in resolving uncertainty about the latent states of a partially observed world, on which agents have complementary perspectives. We then consider the acquisition of the requisite language-entailed by a likelihood mapping from an agent's beliefs to their overt expression (e.g., speech)-showing that language can be transmitted across generations by active learning. Finally, we show that language is an emergent property of free energy minimisation, when agents operate within the same econiche. We conclude with a discussion of various perspectives on these phenomena; ranging from cultural niche construction, through federated learning, to the emergence of complexity in ensembles of self-organising systems.

Charting the landscape of priority problems in psychiatry, part 1: classification and diagnosis

Bayesian inference and posterior probability maps

This paper describes the construction of posterior probability maps that enable conditional or Bayesian inferences about regionally-specific effects in neuroimaging. Posterior probability maps are images of the probability or confidence that an activation or effect exceeds some specified threshold, given the data. Posterior probability maps (PPMs) represent a complementary alternative to statistical parametric maps (SPMs) that are used to make classical inferences. However, a key problem in Bayesian inference is the specification of appropriate priors. This problem can be eschewed by using empirical Bayes in which prior variances are estimated from the data, under some simple assumptions about their form. Empirical Bayes requires a hierarchical observation model, in which higher levels can be regarded as providing prior constraints on lower levels. In neuroimaging observations of the same effect over voxels provides a natural, two-level hierarchy that enables an empirical Bayesian approach. We present a brief motivation and the operational details of a simple empirical Bayesian method for computing posterior probability maps. We then compare Bayesian and classical inference through the equivalent PPMs and SPMs testing for the same effects in the same data.

Analytic measures for quantification of arousal from spontaneous skin conductance fluctuations

Autonomic arousal is often indexed by spontaneous fluctuations in skin conductance. Here, we derive a simple measure of sympathetic arousal, using a convolution model of how sudomotor bursting causes fluctuations in skin conductivity. Under this model, the time-integral of measured conductance is proportional to the frequency and amplitude of sudomotor bursts. We demonstrate the validity of this measure in relation to finite impulse response models, and show that it is a better predictor of autonomic arousal, relative to conventional measures.

Noise-Driven Separation of System and Device Effects in Multimodal Neuroimaging

Simultaneous recordings across spatiotemporal scales using different neuroimagingmodalities are becoming increasingly common. However, this introduces a core ambiguity: do observed differences between signals reflect genuine differences in neural dynamics, or are they artefacts of the distinct ways that devices measure and transform those dynamics? In this study, we address how to disentangle these effects—specifically, how to isolate measurement artefacts introduced by the observer (i.e., the device) from the true dynamics of the system (i.e., the neural system). We show that stochastic fluctuations—typically viewed as a confound — can in fact be exploited when incorporated into generative models. Using synthetic data, we demonstrate that Bayesian model comparison more accurately distinguishes between system- and observer-level differences when models are augmented with noise, compared to noise-free cases. We apply this technique to empirical data and show that cross-scale variation between simultaneously recorded high-frequency broadband signals from macroelectrodes and microwires in the human hippocampus is better explained by differences in observer functions than by underlying system dynamics. These findings suggest that noise, when appropriately modelled, can enhance inference in multimodal neuroimaging by revealing the origin of signal variation across scales.

A mathematical model of embodied consciousness

Perception, as you make it

The main question that Firestone & Scholl (F&S) pose is whether "what and how we see is functionally independent from what and how we think, know, desire, act, and so forth" (sect. 2, para. 1). We synthesize a collection of concerns from an interdisciplinary set of coauthors regarding F&S's assumptions and appeals to intuition, resulting in their treatment of visual perception as context-free.

Bayesian model selection for group studies — Revisited

In this paper, we revisit the problem of Bayesian model selection (BMS) at the group level. We originally addressed this issue in Stephan et al. (2009), where models are treated as random effects that could differ between subjects, with an unknown population distribution. Here, we extend this work, by (i) introducing the Bayesian omnibus risk (BOR) as a measure of the statistical risk incurred when performing group BMS, (ii) highlighting the difference between random effects BMS and classical random effects analyses of parameter estimates, and (iii) addressing the problem of between group or condition model comparisons. We address the first issue by quantifying the chance likelihood of apparent differences in model frequencies. This leads to the notion of protected exceedance probabilities. The second issue arises when people want to ask “whether a model parameter is zero or not” at the group level. Here, we provide guidance as to whether to use a classical second-level analysis of parameter estimates, or random effects BMS. The third issue rests on the evidence for a difference in model labels or frequencies across groups or conditions. Overall, we hope that the material presented in this paper finesses the problems of group-level BMS in the analysis of neuroimaging and behavioural data.

Apathy as a Loss of Prior Precision on Action Outcomes

Abstract Apathy is common in neurological disease, associated with poor prognosis and limited treatments. Current models posit that goal-directed actions are reduced because costs or effort outweigh the expected reward. We highlight an alternative account of apathy, based on the reduction in precision of prior beliefs about action outcomes. In this preregistered study, we test the hypothesis that precision is encoded in the GABAergic gain of prefrontal superficial pyramidal neurons. Fifty healthy adults undertook a goal-directed task during magnetoencephalography. Estimates of synaptic efficacy or gain were obtained by dynamic causal modelling of induced responses. There was strong evidence of a negative correlation between prior precision and apathy (Bayes Factor=12, p&lt;0.01), and that prior precision was associated with gain in prefrontal and premotor neuronal populations (Posterior probability&gt;0.99). The importance of prior precision and GABAergic gain for goal-directed actions opens new avenues to advance the understanding and treatment of apathy.