What can meditative experience tell us about affective valence? A computational perspective on the determinants of affect

Abstract

Understanding affective valence is essential to the fostering of well-being since positive and negative experiences shape our quality of life. One promising approach is to use computational frameworks of perception and action, which seek to formalise the mechanisms underlying valenced experience. Within one such framework, active inference, valence is canonically treated as a consequence of changes in expected uncertainty. This casts valence as a consequence of how effective we believe we are at anticipating phenomena. However, during meditative practice, positive affect is associated with an experiential move into not knowing. If positive valence tracks decreasing uncertainty, then increasing uncertainty should feel worse, yet meditators report the opposite. This suggests that valence may not only depend on changes in uncertainty, and that a more comprehensive account is needed.Buddhism and its meditative practices have long been focused on valence in terms of reducing, and even possibly eliminating, suffering. They seek to achieve this through letting go of the desire to know, claiming that this can feel better than acting on that desire.In this paper I explore what deep and transformative meditative experience can tell us about the computational underpinnings of our valenced experience. I argue this reveals that the complexity of inferential processing plays a key role in determining affect, both during contemplative practice and as a sustained consequence of it. I then introduce a dual-determinant account of valence, in which uncertainty reduction and instantiated model complexity jointly determine experienced valence. This serves to broaden our understanding of valence beyond the current uncertainty-based story.