Agency with economics isn't a big deal

The concept of agency is very useful and important because economics requires economic agents, and intelligence works by economics, so we need economics to describe many common and important phenomena, so we need to be able to see agency in all kinds of unusual places. There are a lot of peculiar connotations that people associate with agency that make it difficult to perceive agency, and we often understand agentic concepts like preference and choice in terms of our first-person experiences with them, which makes it difficult to recognize them in systems where we don’t know what their first-person experiences are like, if they even have any.

A lot of writing on agency has come from philosophers, biologists, neuroscientists, psychologists, and AI researchers. But economists have been working with ideas of agency for far longer that most other sciences—economics is all about economic agents. Even better, agency is basic and simple in economics, making it easy to recognize in many domains. Difficult questions about agency are therefore probably best addressed by referencing economics.

In economics, an agent is basically just a system with preferences and the ability to make choices. A choice is a selection over alternatives. Alternatives are produced within the system as collections of interacting prediction signals that construct motor and cognitive competencies together in conjunction with the environment. These alternatives compete with each other; the winner is propagated across the system such that its anticipated effects occur to the extent that its predictions are correct while losers are inhibited. The selection of a winner is the choice. Thus, many systems that we do not perceive as having first-person experiences can still be seen as making choices from a third-person perspective.

Preferences are the internal factors that bias the construction and selection of alternatives. They multicausally help to produce such outcomes in conjunction with the internal and external environments. Preferences are not written somewhere and held in storage to be accessed when making a decision; instead, they are constructed by the activities of the internal economy. Thus, many systems that we do not perceive as liking things in the same way that we do can still be seen as having preferences.

Thanks to economics (in light of allostasis), we can see that we can study many systems as agents without having to worry about whether they have internal experiences, consciousness, human-like feelings, etc. These are important questions, but they are irrelevant to the question of agency. Agency has been difficult for scientists to study because of the many unnecessary connotations associated with it; hopefully, an economic perspective will allow people to broaden and simplify their understanding of agency, thereby enabling us to discover much more of it.