KT at TSC 2025 Barcelona

A slide-deck presentation that stitches together a theory of consciousness and a clinical brain-stimulation platform, arguing the former should guide the latter.

This is a conference slide deck, so the source is image-heavy and text-sparse. The substantive content is real but the deck's prose is thin; this summary reflects that honestly.

The core bet Ruffini is making is this: consciousness isn't magic, it's what happens when a computational system builds a compressed, predictive model of the world in order to act on it. He calls this Kolmogorov Theory (KT) — named after Kolmogorov complexity, the information-theoretic idea that the "true" description of something is the shortest program that generates it. Brains are special, KT says, because they carry high mutual information with the world in compressed form, and they use that compression to plan. Structured experience — the felt, organized quality of being somewhere, seeing something, wanting something — arises in the act of running those models.

From there, KT makes two moves that matter for psychiatry. First, it defines emotion as model-structure plus valence, where valence is the agent's objective function: how well or badly things are going relative to its goals. Second, it defines depression formally as a pathological state of persistently low valence. That sounds almost tautological, but the point is to give a first-person concept a third-person handle — something you can look for in brain dynamics. The deck cites Ruffini, Castaldo et al. 2024 (Entropy) for the agent-types-of-depression taxonomy, though the details of that taxonomy are not legible from the slides alone.

The clinical half of the talk is about Neurotwin, Neuroelectrics' platform for personalized brain stimulation. The workflow: take a patient's structural MRI and other neuroimaging, build a biophysical head model (predicting where electric fields go in that specific skull and cortex), layer on a whole-brain dynamical model (simulating how neural populations interact), then optimize transcranial electrical stimulation (tES) protocols against that digital twin rather than against a population average. The deck mentions ongoing multicenter trials in epilepsy, major depressive disorder, and Alzheimer's disease, and shows simulation results for psychedelic-state perturbations and parvalbumin interneuron dysfunction (a cellular deficit implicated in Alzheimer's and schizophrenia).

The connective tissue between the theory and the clinic is the claim that KT tells you what you're actually trying to fix — not just a biomarker or a circuit, but an agent's capacity for positively valenced structured experience — and that whole-brain models are the right tool for finding the neural correlates of that capacity. Whether KT does real work here beyond motivating language, or whether it genuinely constrains the modeling choices, is not fully resolved by the slides.