Predicting changes in substance use following psychedelic experiences: natural language processing of psychedelic...

Area of Science:

• Psycho-pharmacology
• Computational Psychiatry
• Behavioral Science

Background:

• Psychedelic drug use (psilocybin, LSD) is anecdotally linked to changes in substance consumption.
• The specific features of psychedelic experiences influencing these changes remain unclear.

Purpose of the Study:

• To investigate if Natural Language Processing (NLP) can quantify psychedelic experiences to predict changes in drug use.
• To determine the predictive power of NLP-derived narrative features for substance reduction.

Main Methods:

• 1141 individuals reporting substance reduction post-psychedelic experience provided narratives.
• NLP topic modeling quantitatively described psychedelic experience narratives.
• Machine learning algorithms predicted long-term substance reduction outcomes using narrative features.

Main Results:

• NLP-derived topic models revealed distinct quantitative descriptions of experiences based on substance class and reduction outcomes.
• Three machine learning models achieved similar prediction accuracy of approximately 65% for long-term outcomes.

Conclusions:

• Machine learning analysis of psychedelic experience narratives shows promise for predicting substance reduction.
• This approach may identify which substances are reduced or quit following psychedelic experiences, aiding therapeutic applications.