Next to context, purpose, means, and decision-making, an agent must have one more thing: choice. An agent must have some alternative action mode available, so that its decision can choose the better of two options. In this way, the agent reaches its objective through a succession of decisions, each of which achieves a relative advantage: the outcome of the chosen option is better than the outcome of the rejected option. Without choice, there is no decision to be made; therefore, a one-trick-pony would not qualify as an agent.
In order to add some depth and colour to this rather abstract sketch, take an archaea as a simple example. In a fit of anthropomorphism, let’s call him Archy. Archy has a unicellular body with so-called flagella on the outside, little hair-like structures that can whip and thus allow Archy to swim around in his watery environment. The flagella give Archy two modes of movement: running and tumbling. Running is literally straightforward, with all flagella paddling in the same direction, thus moving Archy forward. Tumbling is an uncoordinated activity of the flagella, which causes a change of Archy’s orientation. Even though Archy has no means of ‘knowing’ what happens to him (after all, he is just one single cell), these two modes offer him a way of testing alternative routes. He can run a bit, and if the new situation is better than the old (e.g., more food, or less acid), he’ll keep running in that direction. However, if the new situation is worse, he can tumble a little to change course, and then run in this new direction. As long as Archy checks his new situation after every action, and as long as he does not always run (or always tumble), he can influence his fate. Archy has a choice.
Exercising agency is a tall order. It is all the more awe-inspiring to recognise that little Archy masters it within one single cell. Granted: Archy is at the simple end of a spectrum of agents; his example aids conceptual clarity. However, other agents may reach any conceivable level of complexity. There can be several relevant contexts, any hierarchy of missions and objectives, numerous sensors, actuators, and information processors, as well as uncountable choices. All of those may well be contradictory and potentially conflicting; still the agent must integrate them in a reasonably cohesive way to achieve its goal(s). Again: agency is no small feat.
This is the thirteenth in a series of posts on the agency and how it matters to innovation.