Team "Modeling and estimation of subjective states (MESS)"

Research topics

The objective of the MESS team is to uncover the neural and computational mechanisms underlying subjective states such as subjective experience and subjective confidence. We combine invasive human electrophysiology, computational modeling, and translational psychiatry to bridge fundamental neuroscience with clinical applications. The long-term goal is to identify novel diagnostic and therapeutic strategies for neuropsychiatric symptoms such as hallucinations and obsessive doubt.

Axis 1. Perception and Hallucinations

We investigate how the subjective experience associated with a stimulus might emerge from neural activity, and the dysfunction of this mechanism leads to hallucinations. Using single-neuron and local field potential recordings in patients with epilepsy, we dissociate neural correlates of subjective experience from other cognitive processes such as decision-making or attention. Computational models of evidence accumulation and recurrent rate-based neurons are used to explain both veridical perception and induced hallucination-like states. This approach is extended to subclinical psychosis, Parkinson’s disease, or extreme endurance sports, aiming to characterize how over-reliance on priors leads to hallucinations and how this mechanism interacts with neurotransmitter therapy and deep brain stimulation.

Axis 2. Metacognition and Pathological Doubt

We study the mechanisms of confidence (or doubt), and compulsive behaviors. Building on behavioral paradigms and computational models of confidence, we test how metacognitive deficits relate to symptoms of obsessive–compulsive disorder. Patients with obsessive–compulsive disorder perform perceptual decision-making tasks with the possibility of checking, allowing us to model doubt and information seeking. We combine scalp electroencephalography, local field potentials from deep-brain stimulation electrodes, and large-scale online cohorts to identify biomarkers of compulsivity and metacognitive dysfunction.

Techniques used :

• Electrophysiology in humans: single-neuron recordings (tetrodes, intra-operative, Utah array), local field potentials (epilepsy, deep-brain stimulation), scalp EEG.
• Computational modeling: biologically plausible rate-based recurrent neural networks; evidence accumulation and Bayesian models; parameter fitting to individual participants’ data.
• Behavioral paradigms / psychophysics: stimulus detection under uncertainty (visual, auditory and tactile modalities), pseudo-hallucination-induction by expectation manipulation, checking-like tasks, large-scale online cognitive testing.

Partners :