Towards an active inference account of deep meditative deconstruction

This paper explores the intersection of the deep meditative deconstruction resulting from Buddhist meditative practice, specifically through the defabrication process, with the active inference framework (AIF). We propose that defabrication, central to Buddhist meditation, can be understood as a computational process whereby the repeated release of mental tension-associated with clinging and aversion-drives inference progressively lower in an agent's hierarchical generative model. By casting the release of mental tension as a reduction in hierarchical level-specific precision over beliefs about hidden causes, we demonstrate how the phenomenology of defabrication naturally arises from the resultant system dynamics. This approach enables a computational interpretation of Buddhist meditative states, such as the jhānas, and core concepts like equanimity and stillness. More broadly, it advances a computational phenomenology of meditative deconstruction. Finally, we illustrate the deconstruction process under AIF by extending it to the cessation of phenomenal experience, suggesting this offers insight into core-knowledge structuring and the generation of experience.