Language Models Learn Sentiment and Substance from 11,000 Psychoactive Experiences

With novel hallucinogens poised to enter psychiatry, a unified framework for quantifying which changes in consciousness are optimal for treatment is needed. Using transformers (i.e. BERT) and 11,816 publicly-available drug testimonials, we first predicted 28-dimensions of sentiment across each narrative, and then validated these predictions with adjudication by a clinical psychiatrist. Secondly, we fine-tuned BERT to predict biochemical and demographic information from natural language testimonials of drug experiences. Thirdly, canonical correlation analysis (CCA) linked 52 drugs' receptor affinities with word usage, revealing 11 statistically-significant latent receptor-experience factors, each mapped to a 3D cortical atlas. Together, these machine learning methods elucidate a neurobiologically-informed, temporally-sensitive portrait of drug-induced subjective experiences. The models’ results converged, revealing a pervasive distinction between the universal psychedelic heights of feeling in contrast to the grim, mundane, and personal experiences with addiction and mental illness. MDMA was linked to "Love", DMT and 5-MeO-DMT to "Mystical Experiences" and “Entities and Beings”, and other tryptamines to "Surprise", "Curiosity" and "Realization". Applying these models to real-time biofeedback, practitioners could delicately calibrate the course of therapeutic sessions.