Psychedelic-Induced Experiences: New Machine-Learning Models Study Could Help Therapists In Patient Guidance

“Language Models Learn Sentiment and Substance from 11,000 Psychoactive Experiences,” is a research project that took nearly 12,000 drug reports with 52 different molecules and combined them with 3 robotics models to characterize all the trajectories of meaning and feeling as they appear during psychoactive experiences.

The lack of a unified framework for measuring changes in conscious awareness to suit each treatment compelled researchers Sam Freesun Friedman and Galen Ballentine to use two deep transformer-based machine learning techniques for natural language processing (called BERT) and one shallow Canonical Correlation Analysis (CCA) combined with the publicly-available, natural language psychoactive testimonials from Erowid, to try to find answers. 

The first model predicted 28 dimensions of sentiment across each narrative, validated with clinical psychiatrist annotations. A second model was programmed to predict biochemical (pharmacological and chemical class, molecule name, receptor affinity) as well as demographic (sex, age) information from the testimonials. 

Finally, the CCA linked the 52 drugs’ affinities for 61 receptor subtypes with words across the testimonials, revealing 11 latent receptor-experience factors, each mapped to a 3D-cortical atlas of receptor gene-expression. 

Together, the three machine learning methods explain a neurobiologically-informed, temporally-sensitive portrait of drug-induced subjective experiences. 

As the researchers explained, these mutually-confirmatory models point to an underlying structure of psychoactive experience dominated by the distinction between the lucid and the mundane, but also sensitive to effects unique to specific drugs. 

For instance, MDMA was singularly linked to mid-experience feelings of “Love.” Potent psychedelics like DMT and 5-MeO-DMT were associated with “Mystical Experiences.” 

One clear finding is that MDMA is Singular and clearly earning it's title as the love drug pic.twitter.com/PN00DP9ZMk

— Sam Freesun Friedman (@lucidtronix) July 1, 2022

Other tryptamines were associated with an emotional constellation of “Surprise”, “Curiosity” and “Realization”. 

Zooming out to describe the space of psychoactivity at large CCA & BERT found the dominant explanatory factor to spans between mundane suffering in red and lucid beauty in blue. Borrowing poetic license from Yeats we came to call this axis the "Soul Fastened to a Dying Animal" pic.twitter.com/gzahdVuLFh

— Sam Freesun Friedman (@lucidtronix) July 1, 2022

Authors hope that applying these models to real-time biofeedback (like EEG) with zero-shot learning that tunes the sentimental trajectory of the experience through changes in audiovisual outputs will allow practitioners to guide the course of therapeutic sessions, maximizing the benefit and minimizing the harm for patients.

Find the full research here.