Toward aitiopoietic cognition: bridging the evolutionary divide between biological and machine-learned causal systems

We examine and compare autopoietic systems (biological organisms) and machine learning systems (MLSs) highlighting crucial differences in how causal reasoning emerges and operates. Despite superficial functional similarities in behavior and cognitive abilities, we identify profound structural differences in how causality is operationalized, physically embodied, and epistemologically grounded. In autopoietic systems, causal reasoning is intrinsically tied to self-maintenance processes across multiple organizational levels, with goals emerging from survival imperatives. In contrast, MLSs implement causality through statistical optimization with externally imposed objectives, lacking the material self-reorganization that drives biological causal advancement. We introduce the concept of "aitiopoietic cognition"-from Greek "aitia" (cause) and "poiesis" (creation)-as a framework where causal understanding emerges directly from a system's self-constituting processes. Through analyzing convergence pathways including evolutionary algorithms, material intelligence, homeostatic regulation, and multi-scale integration, we propose a research program aimed at bridging this evolutionary divide. Such integration could lead to artificial systems with genuine intrinsic goals and materially grounded causal understanding, potentially transforming our approach to artificial intelligence and deepening our comprehension of biological cognition.