Choice has a context.
Imagine two people being told to take responsibility for their futures. One has a quiet room, reliable transport, savings, a supportive manager and enough time to experiment. The other is caring for a parent, working unpredictable shifts and navigating a workplace where questioning instructions carries a real penalty. The same sentence lands differently in each life. Responsibility matters to both people, but the opportunities to exercise it are not distributed equally. Any serious discussion of agency has to begin there, rather than with a motivational poster pretending the room is empty.
My central argument is that agency is best understood as situated influence: the capacity to select and pursue meaningful actions within conditions that can enable, obstruct or punish those actions. It is not the power to guarantee an outcome. It is not a permanent personality badge. It is a relationship among a person, a set of options, the resources available, and the institutions that decide which options are actually usable. A person can be highly capable and still lack authority. Another can possess authority without exercising sound judgment. Confusing these things makes it easy to blame individuals for systems and systems for every individual decision.
We often talk as though there are only two camps. In one, the heroic individual authors every consequence; in the other, social forces write the entire script. Both can become evasions. The first overlooks luck, dependency and unequal bargaining power. The second can erase the small but consequential decisions people do make: asking for clarification, changing a routine, joining with colleagues, declining an invitation, or revising an opinion after evidence arrives. Those actions may not overturn a structure, but they are not nothing.
The distinction between belief and capacity
Albert Bandura's work on self-efficacy offers a useful psychological distinction. Believing that one can carry out a particular action is not the same as having the external conditions required to succeed. Confidence can help someone initiate and persist with difficult behaviour, but belief alone does not create childcare, fair wages, physical access or a responsive institution. A student might feel capable of learning a programming language and still lack a functioning computer. Conversely, a person with resources may hesitate because earlier failures have made the task seem impossible. A complete explanation has to hold both the subjective and the material in view.
Julian Rotter's locus-of-control framework examines expectations about whether outcomes depend on one's actions or on external forces. It is tempting to turn this into a moral ranking, with an internal orientation automatically celebrated and an external one condemned. That shortcut is unwarranted. In a genuinely arbitrary environment, noticing the role of chance or authority can be accurate. In a setting where effort reliably changes outcomes, recognizing that contingency may be useful. The interesting question is not whether someone sounds optimistic. It is whether their expectations fit the environment they inhabit.
Power changes the menu
Consider a junior employee asked to challenge a senior executive in public. The formal policy may promise open dialogue, yet previous dissenters may have been quietly excluded from important projects. The employee technically has the ability to speak. The practical cost of speaking is another matter. A leader who says, “You should have shown more initiative,” without examining that cost has confused permission with safety. The relevant question is not only whether the employee acted, but whether the organisation made honest participation reasonably possible.
Now consider a family shop where someone wants to introduce digital inventory records. The person may have access to inexpensive software, but the change depends on the routines and trust of relatives who already run the business. They could insist, negotiate, run a limited trial, or abandon the idea. Each option has different relational and financial costs. Here agency looks less like an isolated burst of willpower and more like coordination. The ability to act improves when people can test an idea without turning every experiment into a referendum on loyalty.
Power also creates obligations. A manager can allocate time, determine incentives and protect dissent; a teacher can change what counts as participation; a platform can make a choice easy or deliberately exhausting. Asking only whether users or employees are “taking ownership” obscures the choices made by the people who designed the setting. Accountability should therefore track decision-making power, available knowledge, foreseeable consequences and realistic alternatives, rather than treating everyone as equally positioned.
Accountability without cruelty
There is a genuine moral tension here. If we always explain behaviour through circumstances, we risk making promises and duties meaningless. If we ignore circumstances, we risk punishing people for disadvantages they did not choose. A more careful approach asks what the person knew, what alternatives were realistically accessible, what harms occurred and what repair is possible. This is not a formula that mechanically settles every case. It is a discipline of refusing premature certainty.
Take a missed deadline. Perhaps the worker procrastinated despite having sufficient time; perhaps the brief changed repeatedly; perhaps a medical emergency interrupted the week. The observable event is identical, but the fair response need not be. A responsible supervisor can investigate conditions without declaring the worker innocent in advance. Equally, the worker can acknowledge the missed commitment without accepting a fictional account of complete control. Good accountability is neither automatic absolution nor ritual humiliation.
Philosophy complicates the picture further. The debate about free will asks whether the kind of control required for moral responsibility can exist in a causally structured universe. Compatibilists and incompatibilists disagree over what freedom requires. These debates do not vanish because someone learns a habit or becomes more assertive. The practical language of agency used here does not resolve the metaphysical question. It offers a working vocabulary for evaluating action and institutions while that larger disagreement remains open.
A hypothesis worth testing
Sahil's working hypothesis, not an established research finding: people may exercise agency more effectively when they can distinguish three questions: What can I directly do? What can I influence through relationships or institutions? What must I presently acknowledge as outside my control? The proposed benefit is not relentless optimism. It is better allocation of attention, less misplaced blame and more realistic experimentation. This sounds plausible, but plausibility is not proof. Testing it would require defined outcomes, comparison conditions, repeated observations and attention to who is excluded by the model.
A second hypothesis concerns the language of empowerment. Organisations may get better results when they pair invitations to take initiative with actual authority, resources and protection from retaliation. Yet even that proposal needs careful evaluation: some employees prefer clear boundaries, additional autonomy can increase workload, and participatory processes can become theatre when decisions have already been made. A responsible inquiry should seek cases where the hypothesis fails, not only stories that flatter it.
There are counterarguments. Some will say this framework is too forgiving and risks teaching people to catalogue obstacles rather than act. That danger is real: explanation can become rationalisation. Others will say the focus on individual influence distracts from structural reform. That danger is real too: tiny personal adjustments are no substitute for changing unjust rules. The response is to keep both scales visible. Sometimes the right next move is a personal commitment; sometimes it is collective bargaining, institutional redesign or simply refusing to describe a coercive situation as a free choice.
There is also an epistemic limit. We rarely see the full set of options another person perceived, and hindsight makes alternatives look easier than they were. The fact that someone succeeded under harsh conditions does not establish that everyone could have done so. The fact that someone faced obstacles does not prove that every decision was unavoidable. Our judgments improve when we separate observation, inference and moral evaluation, and when we remain willing to update each one.
Agency, then, is neither a crown awarded to the most assertive person nor a mystical force that bends reality. It is a question we can revisit: Which actions are genuinely available, what shapes their costs, who has the power to change those costs, and what responsibility follows? There is dignity in taking a small action seriously. There is also dignity in admitting that a barrier is real. A mature account of freedom needs room for both.