top of page

What Only You Can Carry: A Definition of Agency

Writer: Christos Makiyama
Christos Makiyama
Oct 17
6 min read

Some years ago, I made a decision I could not fully explain.

The analysis pointed one way. The rewards were clearer, the uncertainty was lower, and everyone reasonable would have chosen it. I chose differently. Not because I knew something others did not, but because something in me, formed over years in ways I could not trace, refused the obvious path.

I could not have justified it properly. I could only say that it did not sit right, and that the not-sitting-right came from somewhere I trusted more than what seemed like common sense at the time.

I have thought about that moment many times since. Not because of how the decision turned out, but because of what it revealed. There was something operating in me that was not knowledge, calculation, or preference. Something that belonged to me in a way that nothing borrowed ever could.

I think that something is agency.

In some of my previous writings, I have described a strange tension. We normally assume that when our capabilities increase, our agency also increases. But this is not always true. Sometimes the opposite can happen. We become able to do more, while something in our ability to originate what we do becomes weaker.

I have argued that agency is worth preserving without clearly defining what I mean by it. The definition matters because we are increasingly being offered substitutes for agency, and we cannot recognize a substitute for something we have never defined.

So I want to attempt the definition here. Not a definition of everything agency may contain, but of what I think sits at its core.

We usually describe agency as the ability to act. This is not wrong, but I think it stops too early.

A machine can act. It can select between options, optimize toward a goal and produce an action much faster and often more precisely than we can. If agency were simply the capacity to act, we would have built it many times over.

Agency is not the capacity to act. It is the capacity to be the origin of an action.

To execute something and to originate it are different things. I can execute a decision someone else made. A system can execute an objective it was given. It can even choose between many possible paths toward that objective. It chooses without being the author of why it is choosing.

So what makes an action genuinely mine?

Freedom alone does not seem enough. Randomness can produce something unpredictable, but I would not call a roll of the dice an agent. Choice is also not enough. A system can choose between options without authorship. Even will does not fully answer it. Will without history is arbitrary.

The answer, I think, is accumulation. Specifically, the accumulation of consequence.

Each of us carries a history. But not only a history of information. Information can be copied, transferred and shared. We also carry a history of consequence.

And that history does not begin with us. We arrive already shaped by what came before us, in our bodies, our instincts and the world into which we are born. Then our own consequences begin to accumulate.

The decisions we made and then had to live inside. The successes and failures that changed us. The relationships that left marks. The losses that rearranged us. The things we did that we could not undo.

These do not only add information to what we know. They change the person who will make the next decision.

There is a difference between learning from a consequence and undergoing it.

When I learn something, I change what I know about the world. When I undergo something, the person who will encounter the world next has also changed.

Someone can know everything about a success or failure I had. I can explain what happened, give them the data, tell them what I learned and even tell them what I would do differently. They can receive the conclusion.

But they cannot receive what it cost me to reach it.

They did not have to live inside that consequence.

This is what I mean by accumulated consequence.

It is not mine because nobody else can experience something similar. Of course they can. It is mine because it belongs to the continuity of my own actions and their consequences. What I did affected what happened. What happened affected me. And the person affected by it becomes the person making the next commitment.

So there is a trajectory, particular to each of us.

I act. There is a consequence. I undergo that consequence. It changes me. And this changed version of me judges and commits again.

Agency exists inside this continuity.

To originate does not mean to begin from nothing. We all begin with what we inherited. What becomes particular to us is what our own consequences make of it.

This continuity also separates agency from capability.

A system can have much more information than I have. It can remember what I have forgotten and recognize patterns I would never see. Increasingly, it can also perform an action better than I can.

None of these capabilities by themselves give it agency.

For much of our experience, capability and accumulation were difficult to separate. Becoming capable usually took time. You became good at something by doing it, failing, correcting yourself and living with the consequences. Capability arrived together with accumulated experience.

Language, money, science and technology have always allowed us to build on capabilities accumulated by others without having to repeat their entire path. What is changing is the scale and speed at which capability can now be acquired without the same process of accumulation. We can receive an answer without struggling toward it, a recommendation without forming the judgment behind it, an action without originating the commitment that produced it.

And I think even the word experience can confuse us here. A system can accumulate experiences as records. But recording a consequence is different from undergoing it. For consequence to matter to agency, what happens through an action must change the continuing agent from which the next commitment will originate.

This is not an argument about something uniquely human.

If one day an artificial system persists through its own actions, undergoes their consequences, is changed by them, and from this changed state originates its next commitments, I could not say that it has no agency simply because it is artificial. That would contradict the definition itself.

The distinction is structural.

Whatever merely processes consequence is not thereby an agent. Agency requires that consequence becomes part of the continuity from which future commitments originate.

Capability is becoming abundant, instant and transferable, while accumulated consequence remains slow, costly and bound to the history of the agent who underwent it.

That abundance is an extraordinary expansion of what we can do.

But it does not expand who is doing it.

And as capability separates from accumulation, we are left with a question we have never had to ask so directly.

If a system can do what I can do, what is left that is mine?

Not what I can do, but where my commitments come from.

This is what I mean by agency.

Agency is the capacity to originate my own commitments from a history of consequences that has become constitutive of who I am. It requires that I remain the author of what I delegate, withhold, or revoke.

Agency does not require that I perform every action myself.

We have always extended our capabilities through other people, institutions, businesses, tools and machines. Delegation itself is not the loss of agency.

The question is whether I remain the author of what I delegate.

Can I reject it? Can I stop it? Can I change direction? Can I take back what I gave away? And when the result comes back to me, does it enter my own process of judgment before becoming my next commitment?

If it does, capability may have been delegated while agency remains mine.

If it does not, delegation stops extending agency and starts replacing it.

This distinction becomes much more important as the systems to which we delegate become more capable. The danger is not simply that machines will act for us. We have built machines to act for us for a very long time.

The deeper question is what happens when they begin to occupy the space between what we undergo and what we commit to next.

Because this is the space where the consequence of yesterday becomes the commitment of tomorrow.

If more and more of that process happens outside us, we may still be the ones who live with the consequences while becoming less involved in originating what follows from them.

This is how agency can contract while capability expands.

Not because somebody takes agency away from us. We may hand parts of it over in exchange for convenience, certainty, speed, capability and even emotional relief. And because the exchange feels like progress, we may not notice the difference until we reach for something that was ours and find that we no longer know how to originate it.

The decision I could not explain years ago was not impressive. It was not wisdom in any grand sense. But it was mine in a way nothing borrowed could be.

It came from somewhere I had built without knowing I was building it.

That somewhere is what I want to learn to protect.

Before we can protect it, we have to be able to see it. And before we can see it, we have to be willing to call it by its name.


The Wounded Angel (1903), Hugo Simberg. Ateneum, Helsinki. Photograph taken in May 2023

bottom of page