Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

Quantifying Psychedelic Visual Phenomena: An AI-Driven Computational Framework for Analyzing Form Constants and Representational Competencies

Cynthia-Maria Kanaan¹, Nadia Mohamed, Rishika Achyuthan, Hannah Demidovich, Jean-Philippe Thivierge; ¹University of Ottawa

Presenter: Cynthia-Maria Kanaan

Psychedelic experiences are widely reported to induce vivid visual phenomena characterized by recurring geometric patterns, known as form constants (Bressloff et al., 2002). These hallucinations have been shown to provide therapeutic benefits (Aynsworth et al., 2017; Gattuso et al., 2022; Leptourgos et al., 2020) and offer insights into the neural mechanisms of altered perception, through their ability to disrupt normal states of consciousness (Carhart-Harris et al., 2016; Greco et al., 2025; Suzuki et al., 2017). However, the processes underlying these effects are not yet fully understood as existing approaches lack a standardized, quantitative framework to characterize and study them (Castelhano et al., 2021; Gattuso et al., 2022). This study addresses these limitations by introducing a proof-of-principle methodology that leverages generative artificial intelligence (AI) to extract, quantify, and classify perceptual features commonly associated with experiences of visual hallucinations. A generative pre-trained transformer was developed to generate image-to-text descriptions focusing on underlying form constants and their mathematical properties, while avoiding surface-level content. The custom model was employed to analyse 40 psychedelic artworks inspired by subjective experiences (i.e., containing hallucinatory patterns), and 40 non-psychedelic images collected from online sources. Texual outputs were converted into semantic embeddings using a transformer-based encoder in the R language. Hierarchical clustering and cosine similarity analyses revealed strong within-group and low between-group similarity for both categories (Figure 1). Principal component analysis (Figure 2) further confirmed the AI’s ability to capture distinctive themes characteristic of psychedelic imagery, demonstrating how the proposed computational method can systematically analyse subjective visual phenomena. This contributes to bridging the gap between subjective psychedelic experiences and objective quantitative analysis, offering a novel tool for understanding neurological and perceptual mechanisms underlying altered states of consciousness. It lays the groundwork for future research into how artificial systems may simulate and process high-dimensional sensory data, advancing cross-system representational competencies.

Topic Area: Methods & Computational Tools

Extended Abstract: Full Text PDF