Entropy (Basel, Switzerland)
January 19, 2025
Giulio Ruffini, Francesca Castaldo, Jakub Vohryzek
6 citations
Tracking natural data forces an agent to mirror the symmetry properties of the generative world model, enforcing a hierarchical organization in the agent's neural network consistent with the manifold hypothesis. Using Lie pseudogroups to formalize invariance in natural data and drawing parallels to Noether's theorem, the study shows that data tracking constrains both the agent's constitutive parameters and dynamical repertoire. This bridges algorithmic information theory, symmetry, and dynamics, offering insights into neural correlates of agenthood and structured experience, as well as AI and brain model design.
Zenodo (CERN European Organization for Nuclear Research)
June 28, 2026
Giulio Ruffini, Francesca Castaldo
Pharmacological neuroplastogens like psilocybin and LSD enhance neural plasticity by flattening high-level priors, allowing bottom-up prediction errors to remodel the brain's generative model. The same computational regime can be achieved non-pharmacologically through immersive algorithmic art held in a Goldilocks zone of compressibility. This approach is operationalized in a closed-loop digital therapeutic for adolescent depression. The argument extends to music, where harmonic tension serves as a prediction-error scaffold, and live performance with a chaos-harmony narrative arc. All three modalities sustain structured prediction error in the Goldilocks zone, transiently flatten the dynamical landscape, and push subjective phenomenology into territory typically associated with psychedelics like MDA, psilocybin, and LSD, as measured by altered states of consciousness and mystical experience instruments.
Zenodo (CERN European Organization for Nuclear Research)
June 28, 2026
Giulio Ruffini, Francesca Castaldo
Immersive algorithmic art may enhance neural plasticity through the same computational mechanism as psychedelics: sustained, structured prediction-error signaling. The brain's modeling engine generates predictions of sensory input; mismatches drive model updating via synaptic plasticity. Algorithmic art maximizes these errors while keeping stimuli in a compressible, emotionally rewarding "Goldilocks zone," creating a self-reinforcing loop of engagement, prediction error, plasticity, model updating, and positive valence. The hypothesis is formalized within Kolmogorov Theory, connected to the REBUS model, and supported by convergent evidence from psychedelic neuroimaging and predictive-coding electrophysiology. A translational pathway combining closed-loop EEG-driven algorithmic art with cognitive behavioral therapy for adolescent depression is outlined.