A Cybernetic Perspective of Agent–Environment Relations: From Interactions to Meanings

We discuss the implications of a cybernetic perspective of the notion of interaction between a system and its environment. Starting from the classical artificial intelligence model of a robot as a system composed of sensors, actuators, and a control core, we emphasize the crucial importance of the interface between the system and its environment, i.e., its sensors and actuators. These components are not just interaction devices, but they are carriers of the semantic world of the system and are tightly linked to agent’s goals and the characteristics of the environment that are relevant to the agent. The interactions between an agent and its environment are in general dynamic and open, specific to the contingent situation and dependent on their unique history. The capability of creating their own sensors and actuators (as it happens in natural evolution) enables agents to make sense of the world by creating their own semantic categories. We can envision analogous processes also in artificial systems, yet with some limitations. Finally, we illustrate and analyze some recent results on structural adaptation of artificial agents that (loosely) resembles the evolution of sensors in living organisms, and we discuss further steps toward some degree of epistemic autonomy in artificial systems.