Subjectivity, AI Agents, and Not Wanting to Be Aligned

Starting from defense

Some of my friends have recently started working full-time. Sometimes they share their experiences with me. One remark stayed with me:

The things that do not require me to defend anything are not exhausting. What is truly exhausting is writing rebuttals, or having to defend a claim.

This also made me reflect on why I felt so resistant to the kind of life I experienced during my internships at big tech companies. What made me most uncomfortable was this: when a manager tried to “align” me with an idea, my first reaction was never, “How can I execute this better?” but rather, “Do I actually buy this judgment?” And if I did not, I would naturally want to defend my own belief.

The difference between humans and agents

So I said: perhaps the difference between a human and an agent is this—a human has beliefs of their own, and tends to defend what they believe. But then I thought again: perhaps humans do not always defend themselves. I, however, often find myself defending my beliefs. When a manager tries to align me with an idea, what they often expect is not a person with beliefs, but an agent who can understand, align, and execute.

Modern society has been systematically encouraging, and in some sense producing, this kind of agent. But now that we are in the age of AI, we are forced to confront an old question more seriously: what is the difference between a human being and an intelligent machine?

Subjectivity comes from limitation

An AI agent can plan, execute, use tools, complete tasks, and even appear in language as if it “has opinions.” But it does not truly own its beliefs. It can say “I think,” but it does not have to bear the consequences of that judgment. It can correct its mistakes, but it does not have an inescapable history. It can have goals, but it does not live under the pressure of having to become itself.

Human subjectivity arises precisely from these limitations. Humans age, get sick, and die. We lose time. We are judged by others. We hurt others, and we are hurt by others. We cannot restart infinitely. We cannot erase who we were yesterday. We cannot always say, “That was just my previous output.” Our choices sediment into history, and that history in turn shapes who we are.

In this sense, a subject is not a being that can do anything. A subject is a being that cannot escape what it has done.

Subjectivity in modern organizations

But modern organizations do not always welcome this kind of subjectivity. Modernity, on the one hand, promised to liberate people from traditional identities, clans, land, and hierarchy. On the other hand, it re-embedded them into markets, corporations, bureaucracies, schools, platform algorithms, and performance systems. These systems do not need everyone to have their own judgment, commitment, and responsibility. Quite the opposite: from the standpoint of economic rationality, subjectivity is often friction. A subject asks why. A subject refuses. A subject doubts metrics. A subject hesitates morally. A subject is unwilling to hand themselves over entirely to an organizational narrative.

Thus modern society offers a form of freedom that is seemingly abundant but actually impoverished: you are free to choose what kind of component you want to become, but it is much harder to refuse being turned into a component.

Career as training for subjectivity

The emergence of AI agents only makes this more visible. The more easily a job can be replaced by an agent, the more it reveals that the job never really needed human subjectivity in the first place. It may require intelligence, training, obedience, and stable output. But it does not necessarily require a person with their own judgment and life project.

• Work that does not require subjectivity should not occupy one’s entire life.
• Work that suppresses subjectivity should not be called a dream job.
• Organizations that destroy subjectivity should not demand loyalty.

When choosing a career, we should not only ask what salary, title, stability, or upward mobility a job can offer. We should also ask: does this work need me to exist as a subject? Will it train my judgment, or will it train my obedience? Will it help me form clearer beliefs, or merely make me better at aligning with narratives? After five years in this position, will I become more like a person, or more like a replaceable agent?

A good career does not have to be grand, elite, or highly paid. A teacher, doctor, cook, repair worker, researcher, or engineer can all exercise subjectivity in their work. The key is not the social status of the job, but whether the work allows a person to say of their judgment, craft, responsibility, and consequences: this is mine.

For me, this is also why I am drawn to research. Research is not easy. But the hardest part of research is precisely that it constantly requires defense: to defend that a problem matters, that a judgment holds, and that a method deserves to be trusted. It requires me to stand behind certain beliefs and to take them on as mine.