Thoughtseeds: A Hierarchical and Agentic Framework for Investigating Thought Dynamics in Meditative States.
Entropy (Basel, Switzerland) April 24, 2025 Prakash Chandra Kavi, Gorka Zamora-López, Daniel Ari Friedman et al. 2 citations
A computational model called the Thoughtseeds Framework simulates thought dynamics during focused-attention Vipassana meditation. It treats thoughts as dynamic attentional agents organized in a hierarchy of nested Markov blankets across three levels: knowledge domains, a network where thoughtseeds compete, and meta-cognition that regulates awareness. The model generates four states—breath control, mind wandering, meta-awareness, and redirecting to breath—through self-organizing interactions. Expert meditators sustain control dominance, reinforcing focused attention, while novices show frequent and prolonged mind wandering, reflecting instability. The framework integrates Global Workspace Theory and active inference to explain how meta-awareness shapes a unitary meditative experience, offering testable predictions about meditation skill development.