Quantitative measures for autonomy and emergence, grounded in Granger causality and multivariate autoregression, are introduced and validated. G-autonomy quantifies how much a variable's past predicts its own future beyond external factors, while G-emergence measures a process's simultaneous dependence on and autonomy from its underlying causes. Applied to agent-based models, evolutionary adaptation increases autonomy in a predation model, and a flocking model demonstrates both emergence and downward causation. The work connects these measures to broader discussions of autonomy, emergence, and consciousness.
Dualisms between mind, body, and nature have shaped cognitive science and AI, but advanced AI systems that create art or music prompt calls for a shift in values toward AI ethics, rights, and personhood. While discussing agency and rights is not wrong in principle, it misdirects attention in current circumstances. Questions about artificial agency can only follow a genuine reconciliation of human interactivity, creativity, and embodiment. The authors contribute to embodied and enactive approaches to AI by exploring interactive and contingent dimensions of machines through Japanese philosophy. A key takeaway is that AI and machine learning systems should be recognized as powerful tools or instruments, not as agents themselves.
A workshop at the Artificial Life XV conference revisited enactivism's contributions to biology, aiming to ground the concept of autonomy in quantitative definitions based on observable phenomena. Discussions covered identifying emergent individuals from environmental backgrounds, the roles of autonomy and normativity in biological theory, the spontaneous emergence of autonomous agents at life's origins, and scientific approaches to subjective experience.
Using simulated agents, the authors model how an embodied agent and its environment influence each other via a sensorimotor loop. Information-theoretic measures quantify information flows, including morphological computation (interaction between body and environment) and controller complexity, which relates to integrated information theory of consciousness. Prior work found that a well-adapted morphology reduces needed controller complexity. Here, the authors observe that agents must first understand relevant environmental dynamics to interact effectively, so increased controller complexity can improve body-environment interaction.