PERSON-SPECIFIC NEUROMODULATION
Synthetic research model
No patient data or demonstrated therapy effects
INDEPENDENT TASK READOUT
A readout trained on a separate reference session estimates a simple input signal. These scores are engineering outcomes, not measures of autistic wellbeing.
REPRODUCIBLE COMPUTATIONAL EXPERIMENT
24 synthetic individuals. Three evaluation seeds. Six strategies. A separate reference session for each individual.
| Strategy | Input tracking error | Decoding accuracy | Control energy |
|---|
Means and 95% bootstrap intervals across synthetic individuals. Seeds are averaged within each individual. These intervals do not describe patient treatment effects.
Whether a specified feedback rule works in a known synthetic dynamical system, and where it fails. It cannot establish clinical efficacy, stimulation safety, durable brain rewiring, or an autism-specific entropy target.
Neural dynamics: Wilson–Cowan-type excitatory and inhibitory populations. Plasticity: bounded rate covariance with relaxation toward baseline. Observation: a noisy population average, not scalp EEG. The anatomy is a display surface; the coupling matrix is synthetic.
EVIDENCE THROUGH 7 SEPTEMBER 2026
Three distinct questions: how brain dynamics differ, whether feedback changes those dynamics, and whether the change helps a person.
In 62 adults, psilocybin-related brain trajectories were aligned with the sensory context. This challenges the idea that increased signal diversity is simply unstructured noise. It is not evidence of benefit in autism.
Stoliker et al. · Nature ↗A study of 1,824 people found very little overlap in extreme deviations at individual connections, but more convergence at regional and network scales. Atypical does not automatically mean harmful or identify a treatment target.
Ilioska et al. · Nature Mental Health ↗A sham-controlled trial in 200 autistic children reported better social-communication scores after motor-cortex stimulation. Follow-up lasted one month. This was neither an entropy-targeted nor a closed-loop treatment.
Tan et al. · The BMJ ↗Two trials of different frontal targets did not show superiority to sham on their primary autism symptom outcomes. Target, population, protocol, and endpoint matter.
Ni et al., 2023 ↗ · Ni et al., 2024 ↗In one child EEG study, higher generalized entropy coexisted with lower Lempel–Ziv complexity. This lab specifies its estimator and tests measurement noise.
Tenev et al. · Frontiers in Psychiatry ↗A four-person randomized feasibility study of adaptive deep brain stimulation in Parkinson’s disease supports the engineering approach. Its benefits cannot be transferred to autism without direct evidence.
Oehrn et al. · Nature Medicine ↗Can a reliable, person-specific complexity measurement improve a consented functional outcome beyond simpler feedback, under realistic observation and actuation constraints?
The research manuscript and downloadable Python package define the equations, comparisons, data provenance, and limitations. No implant, drug administration, or human stimulation is performed by this application.