Conference Agenda

Background. Patients with anorexia nervosa (AN) typically hold aberrant beliefs about their body that they struggle to update. These can vary from sensory, local beliefs about how full their stomach feels to global, prospective beliefs about how much they can trust their own body. Questionnaire studies revealed the role of such beliefs in the onset, maintenance, and treatment of AN, yet the parameters that contribute to their formation and maintenance remain unknown.

Aims and Methods: I present two studies with independent samples at the acute (total N = 106) and post-acute AN state (N = 113), compared to matched healthy controls (N = 220) that aimed to: (1) assess interoceptive and exteroceptive perception and metacognition using state-of-the-art methods and statistical approaches such as the Heart Rate Discrimination Task (Legrand et al., 2022; M ratio analyses); (2) develop and validate a novel neurocomputational approach using a Bayesian Learning Framework to determine the parameters that may influence explicit belief formation and updating; (4) relate some of these parameters to clinical symptoms such as insight into illness, using measures validated specifically in eating disorders.

Results and Conclusions. AN patients showed lower precision-weighted, learning rates than controls in both retrospective and prospective belief updating across domains. However, exteroceptive and interoceptive metacognition, and key, latent parameters of belief updating, predicted different disruptions of clinical insight; exteroceptive, local belief updating difficulties were linked with unrealistic body image beliefs, while difficulties in interoceptive metacognition predicted global insight difficulties.

Social neuroscience has studied interpersonal dynamics using neural and behavioral synchrony markers (Schilbach & Redcay, 2024). The 5E cognition approach (Embodied, Enactive, Extended, Embedded, and Ecological) - historically called generative neurophenomenology approach (Varela, 1999; Monnier et al., forthcoming, 2025) links bodies synchrony with lived experiences of “interbeing”, such as empathy and togetherness (Troncoso et al., 2023). Applied to Autism Spectrum Disorder (ASD), these dimensions help explain atypical social interactions, particularly through the interpersonal misalignment model (Bolis et al., 2022). However, these aspects remain underexplored together, especially with autistic children.

This study simultaneously collects and analyzes the dynamics of neural, cardiac, motor, and subjective synchronies on 80 Mother-Child dyads during cooperation tasks, including 40 verbal ASD children. Electroencephalography (EEG) and electrocardiography (ECG) measure simultaneous activity in a hyperscanning setup, while cameras capture motor synchrony. The "Inclusion of Other in the Self" scale (Aron & Aron, 1992) assesses perceived togetherness after each activity. Body maps and microphenomenology interviews provide additional qualitative data on lived experiences.

Initial results from a pilot study of 10 dyads reveal correlations between interbrain synchrony and sense of togetherness. We explore alpha-mu rhythm desynchronization (8-12 Hz) in the right temporoparietal junction as a marker distinguishing ASD from neurotypical children. Motor coordination differences are investigated to predict interbrain synchrony and the level of shared experiences.

This project fundamentally aims to characterize the dynamics of physiological and subjective experiences in social interaction. Clinically, it challenges the deficit-based view of autism, by offering unique multimodal interpersonal data and a reusable Python toolbox.

The term Peripersonal Space (PPS) defines as the space immediatelly sorrounding the body, where the body can physically interact with stimuli in the external enviroment. In this context, PPS has been described in multiple ways, including reaching space, defensive space, interpersonal space, action fields, and contact prediction mechanisms, or the “space of the self”. Here I will to discuss the fundamental nature of PPS as a neural representation integrating tactile inputs from the body with external cues that have the potential to contact the body —i.e., stimuli within near space. In this context, while PPS has traditionally been framed in spatial terms, I propose that its core function extends beyond spatial computations. Instead, PPS encodes the contingencies between bodily and external stimuli in both space and time. This shift in perspective suggests that PPS should be reconceptualized as Peripersonal Space-Time (PPST). In this context, PPST grounds a basic representation underrlying self-consciousness “here and now”. While direct empirical validation of this hypothesis is still forthcoming—experiments are currently ongoing —there is substantial indirect evidence supporting this framework. I will outline a roadmap for future research to assess and refine this proposal.

The sense of agency (SoA)—the experience of controlling one’s actions and their effects—is essential to consciousness. Traditional SoA studies rely on explicit responses, limiting assessment in those unable to provide feedback, such as infants, non-verbal individuals, and artificial systems.

We propose that active causal inference of one's causal power serves as a behavioral signature of SoA. Our study examined how individuals infer their causal power through structured action. Using transformer-LSTM-based deep neural networks, we quantify the connection between active inference behaviors and SoA.

We conducted control detection and judgment tasks with adults, children, and individuals with schizophrenia, who used a mouse or touchpad to manipulate visual stimuli. We analyzed movement trajectories using deep neural networks to quantify action plans and examined how action plan diversity related to control detection accuracy.

Findings indicate that action plan diversity systematically increases with perceived control, reflecting an active exploratory strategy for accumulating causal evidence. In adults, action plan geometry predicted behavioral responses in two control tasks, confirming the role of actively structured action policies in agency computations. In children, action plan diversity correlated with developmental improvements in agency perception. Individuals with schizophrenia exhibited reduced action plan diversity and impaired agency inference, indicating limited capacity for causal evidence gathering.

Our research demonstrates that active causal inference plays a crucial role in SoA. We argue that it serves as a signature of SoA. By observing active causal inference behavior without requiring explicit reports, our study can potentially establish the first no-report paradigm for objectively assessing SoA.

The Rubber Hand Illusion (RHI) is an important paradigm in consciousness research, offering insights into how the brain constructs a coherent sense of body ownership and self-awareness. To investigate the integration of multimodal sensory information and the brain’s ability to form a unified self-representation, we developed a neuro-computational model that mimics the role of the parietal cortex in the RHI experiment.

Our multisensory integration network learns the joint distribution of sensory inputs by observing different arm positions through proprioception and vision, encoded in separate populations that converge upon a multisensory population with recurrent connections to an inhibitory population. Unlike previous models, our approach leverages purely local Hebbian learning to achieve human-like multisensory integration. Under normal conditions, our network achieves Bayesian-optimal decoding accuracy of true arm positions. In the RHI condition, we observe a proprioceptive drift pattern similar to human behavior: for low sensory disparities (<25°), the model fuses both inputs, leading to intermittent arm position estimates. For higher disparities, the model exhibits a winner-takes-all behavior, where the more reliable input dominates the sensory estimate. Thus, our model generalizes to out-of-distribution inputs, demonstrating human-like patterns of proprioceptive drift consistent with Bayesian causal inference as observed in human data. Additionally, we validated our model in a robot-hand illusion experiment using an iCub robot, demonstrating that robots can develop a body schema purely through unsupervised sensory learning. Our model provides a mechanistic explanation of how flexible body ownership emerges and offers a computational foundation for consciousness grounded in bodily experiences.

The embodied experience of emotion has been discussed for centuries (James,1884; Schachter & Singer, 1962), with some describing the perception and interpretation of interoceptive bodily sensations as integral for emotion. This translates in dysfunctional mental health cases where bodily processes are often misrepresented. This visceral misrepresentation influences affective processes through both conscious and unconscious pathways.

Therefore, I will present findings from a variety of methods which capture both conscious and unconscious bodily influences on mental health and affective experiences. This will include psychophysical tasks of assessing cardiac and respiratory interoception, multidimensional interoceptive experience sampling of embodied thoughts, and methods which assess implicit interoception such as via electrophysiological coupling between neural and visceral rhythms.

We observe distinct mental health relationships between objective and subjective perception of interoceptive signals. Specifically, dysfunctional interoception in mental health symptoms may demonstrate more predominantly in conscious subjective experiences rather than objective sensing of interoceptive signals. This is demonstrated with broad mental health associations with both subjective survey and experience sampling methods of assessing interoception, but not with interoceptive task performance measures. In addition to complex conscious interoceptive influences on mental health, we also observe unconscious contributions of interoception (i.e., stomach-brain coupling) with multidimensional symptoms of mental health.

These findings highlight the multifaceted and complex role of interoception in mental health, emphasising the importance of capturing broad conscious and unconscious interoceptive processing. Relying solely on performance-based tasks risks overlooking critical interoceptive contributions from subjective and implicit measures, essential for a comprehensive understanding of interoception on mental health.