Active inference as a theory of sentient behavior

Epistolution: a new principle necessary to a learning-first theory of life

“…The study of self-organization is a space where an alternative telos of life is continually proposed. Even though they not been explicitly recognized and celebrated as such, ideas such as Kaufmann’s “order for free [ 28 ], ” Friston’s free energy principle and active inference [ 29 ], Branscomb’s anti-entropic machines [ 30 ], and others, are conjectures about what is prior to life, about what process allows for inheritance, and therefore about the telos of life. These theories, masquerading as molecular chemistry and physics, are also guesses about fundamental motivations, the purposive goals that lead life to prefer order and purge entropy.…”

RDoC Framework Through the Lens of Predictive Processing: Focusing on Cognitive System Domain

“…What is unique about the PP framework, is its power to formulate hypotheses (Pezzulo et al, 2024). These hypotheses make specific empirical predictions that span different units of analysis, ranging from gene to behavior, which can be empirically validated.…”

“…At the physiological level, PP has empirical predictions about the top-down and bottom-up dynamics supporting prediction and prediction error under the name of predictive coding (K. Friston, 2018). Additionally, it elucidates how motor functions (e.g., oculomotor performance) and higher-level cognitive processes (e.g., planning) manifest as an active form of prediction error minimization, under the name of active inference (Parr & Friston, 2018;Pezzulo et al, 2018Pezzulo et al, , 2024.…”

“…At the circuit level, the exchange of information between top-down and bottom-up pathways can be understood through oscillatory dynamics within and across brain areas, potentially indicating temporal predictions (Bastos et al, 2012;Walsh et al, 2020). On the molecular and cellular levels, synaptic activity and efficacy, modulated by neuromodulators, correspond to inferential processes that minimize free energy across faster and slower time scales, respectively (K. Friston, 2023;Parr & Friston, 2018;Pezzulo et al, 2024). This involves precision dynamics, which balance inferential processes at multiple levels by adjusting the post-synaptic gain of sensory or prediction error units (K. Friston, 2023).…”