What machine learning tells us about microdosing

Natural language signatures of psilocybin microdosing

The benefits of microdosing can be so subtle that they’re hard to measure in clinical settings, which is why this study turned to machine learning and natural language processing.

Essentially, patients were asked a series of subjective questions following psilocybin microdoses or placebo doses, and the interviews were analyzed with AI.

The key takeaways?

  1. A “sentiment analysis” measured the connotation of each word as positive, negative, or neutral. Sentiment scores increased while microdosing, suggesting that psilocybin has a positive effect on mood and subjective well-being.
  2. Verbosity (length of responses in number of words) increased for all questions when microdosing, which may reflect increased enthusiasm, motivation, and energy.
  3. Semantic variability (how much the meaning of consecutive words changed) was not affected by microdosing. This implies that microdosing does not lead to incoherent speech and might not enhance creativity like many believe. 

It’s important to note that the microdose were fairly large – 500 mg of dried mushrooms. It would be interesting to see how a 100 or 200 mg dose compares.

PDF of article