An agent is an autonomous entity with the capacity to perceive its environment, process that information, and act based on that perception. Agents are the active “actors” of any complex system.
An agent does not need to be intelligent or conscious. A cell, an ant, a human, an institution, or a local software loop are all agents. What matters is that their actions are never completely random; they are always a response to their internal state or their immediate surroundings.
Furthermore, agency is not just about the ability to act, but also the capacity to not act. Inhibiting a default, automatic response is often the most resource-intensive aspect of agency. Under stress or overload, this capacity degrades, and agents naturally fall back to rigid, automated patterns.
In a complex system, agents are often nested: muscle cells are agents that form an organ, which is part of an athlete, who is an agent within a sports team. No single agent has a map of the entire system. Instead, global organization emerges from the bottom up as agents respond to what they perceive, often coordinating indirectly through stigmergic traces.