A person sees that a public-service decision is wrong. They understand the evidence and know how to explain the error, but the form offers no appeal and the office gives them no authority to change it. Calling this person “empowered” because they were asked to click a confirmation button would mistake a gesture for an ability to act. Agency concerns the capacity to form purposes, make choices and affect what happens within real conditions. It includes individual action, action through others and collective action. It is shaped by knowledge and motivation, but also by time, rights, resources, relationships and institutional power. This entry gives a research-led definition that keeps those dimensions together. It also separates agency from self-efficacy, autonomy and the sense of agency, terms that sound similar but answer different questions.
Agency is more than the feeling of choosing
Albert Bandura’s social cognitive account describes agency through intentionality, forethought, self-regulation and self-reflection. People can imagine an outcome, select a course, adjust their behavior and evaluate what happened. They are influenced by social structures while also acting within and sometimes changing those structures. This is not a claim that every individual is unconstrained. Bandura’s review explicitly places personal agency in a network of sociostructural influences. A person may have a goal and a plan yet lack the material or political conditions to carry it out.
Bandura distinguishes direct personal agency from proxy agency and collective agency. A patient may act directly by choosing among understandable treatments. They may act through a representative who can obtain a needed service. Neighbors may organize to change a dangerous crossing that no one household can alter alone. These modes remind us that agency is not identical to solitary independence. Relying on another person can be an expression of agency when that person is accountable to the individual’s purpose. Collective action can extend options beyond what isolated people can achieve.
A moment of choice is therefore insufficient as a measure. An interface might offer three buttons, all leading to the same outcome. A workplace might invite suggestions without allowing staff to change the procedure. In those cases the appearance of participation may exceed the effective range of action. Conversely, a person who delegates a task after informed consideration can remain an agent even when they do not perform every step themselves. The relevant question is what options were actually available and whether the person could use them to pursue a valued end.
Past, future and present in action
Sociologists Mustafa Emirbayer and Ann Mische describe agency as a temporally embedded process. Past habits supply familiar ways of acting; future-oriented imagination opens alternatives; practical evaluation responds to the contingencies of the present. This three-part account adds something to a purely individual skill model. A person does not decide outside history. They inherit roles, routines and expectations, while also imagining what could be different and judging which move is possible now.
Consider a teacher facing a new digital assessment system. Habit may suggest entering scores as before. A future goal may be to make feedback more useful to students. The present situation includes a deadline, school policy, software limits and students’ needs. Agency appears in how the teacher works across these times: drawing on what has worked, imagining an improvement and making a situated decision. A measure that records only whether the teacher clicked “accept” would miss the temporal process and the constraints that shape it.
The sociological account does not replace Bandura’s psychological account. It asks a different level of question. Bandura emphasizes mechanisms of intentional action and self-regulation; Emirbayer and Mische emphasize the relation between action and social structures through time. Together they help prevent two mistakes: treating people as passive products of context, and treating context as irrelevant once a person has been told to choose. Both forms of explanation need evidence at the level they claim.
Self-efficacy is related but narrower
Self-efficacy is a person’s belief that they can organize and execute actions for a particular task or situation. It can affect whether they begin, persist and recover after difficulty. Agency is broader. A person may be confident that they can complete a form yet lack permission to submit it. Another may doubt their ability, seek help and successfully act through a trusted advocate. A general statement that someone “has agency” cannot be inferred from one confidence rating.
Belief matters because it can help a person attempt a feasible action. Bandura’s earlier self-efficacy research showed how expectations of capability relate to behavior in specific contexts. But a belief cannot create an absent opportunity. Training that increases confidence without changing an inaccessible service may produce frustration rather than effective action. A robust analysis measures both the person’s task-specific expectations and the real affordances of the environment. It also asks whether new action produces an outcome the person values.
The distinction is useful in education. A student may believe they can revise an essay, understand the feedback and have enough time to do so. Another may believe the same but lack a quiet place or access to the draft. A third may have resources but be unsure where to begin. Supports differ across these cases. We should not collapse them into a single “agency score” that hides whether the obstacle is knowledge, confidence, time, authority or discrimination.
Autonomy is not the same as acting alone
Self-determination theory distinguishes autonomous motivation from controlled motivation. A person can act autonomously when they endorse the reason for an action, even if they cooperate with others or follow expert guidance. They can also act alone under pressure or fear. Deci, Olafsen and Ryan’s workplace review describes autonomy alongside competence and relatedness as psychological needs that shape motivation and well-being. This is a useful neighboring lens, not a complete definition of agency.
Agency asks about forming and pursuing purposes with effect. Autonomy asks, in this tradition, about volition and endorsement. A worker can choose a task that fits their values yet lack authority to change the resources needed to complete it. They may have substantial operational power but feel compelled to pursue a goal they reject. These are different problems. A manager who offers a menu of tasks should not assume that autonomy or agency has been secured without examining pressure, information and consequences.
Relationships can support both. Good guidance can increase a novice’s range of action; a colleague can help someone use an appeal process; a community can sustain a project that would otherwise fail. The opposite is also true: dependence can become coercive when the helper controls access or withholds information. The question is not whether another person is involved. It is whether the arrangement expands or narrows the individual’s meaningful options and endorsed participation.
Sense of agency is an experience, not a capacity
Sense of agency names the feeling of controlling one’s action and its effects. Haggard’s neuroscience review examines how that experience may arise from action-related signals and feedback. It is important, but it does not measure the whole social capacity described here. Someone can feel in command while an interface quietly constrains the options. Someone can feel little immediate control during a long collective campaign while the group is in fact changing policy. Experience and effective influence can diverge.
This distinction matters in technology. A system that gives fast feedback may make users feel responsible for a recommendation they could not inspect or alter. A slow, complex process may make a user feel detached even when their informed choices have real effects. Neither feeling alone proves the allocation of power. Product teams should test the subjective experience and separately audit decision rights, reversibility, access to evidence and appeal. The Institute’s proposal is to keep these measures beside each other, not to combine them into a single flattering label.
The existing Library page on sense of agency covers that experiential construct directly. This article focuses on the broader capacity to act, including its social and institutional conditions. The crosslink is deliberate: readers should be able to move between the two without assuming that one is a synonym for the other.
How agency can be observed
Agency is hard to reduce to one measure because it spans purpose, options and effects. A researcher might interview people about goals and constraints, observe whether choices can change an outcome, or analyze institutional rules that distribute authority. Each approach sees something different. A survey asking “Do you feel in control?” measures a perception. A log of button clicks measures a behavior. An audit of who can reverse a decision measures an institutional condition. A full account should explain which level is being studied and why.
In a school, one could examine whether students can choose questions to investigate, obtain feedback and revise work. In a clinic, one could ask whether consent options are understandable, whether alternatives are accessible and whether a patient’s preference changes the plan. In a public service, one could test whether an appeal is usable by the people most affected. These are examples of possible operationalizations, not a universal validated battery. The correct measure depends on the decision and on whose agency is at stake.
Outcomes also require care. An agent can make a considered choice and still fail because circumstances change. A successful outcome can occur without meaningful participation, as when a default happens to align with what someone wanted. Agency is not identical to success. It concerns the process and real possibility of pursuing an end, including the ability to revise after feedback. A sound evaluation therefore looks beyond whether the final result was favorable.
Agency in social and digital systems
Institutional design can enlarge or shrink agency. A form with plain language can help a person understand options; a clear deadline can make action feasible; a right to appeal can correct errors. But a polished interface cannot substitute for substantive authority. If eligibility criteria are hidden, if there is no reachable human reviewer, or if a deadline is impossible for some users, the system may make agency appear greater than it is. The design question is whether a person can understand, choose, act and contest outcomes in time to matter.
AI-assisted decisions sharpen this issue. A caseworker may receive a recommendation with a confident score. If policy requires accepting it, the human click is not meaningful oversight. If the caseworker can inspect evidence, seek a second view and change the result, their agency is more substantial. The affected person also needs a route to know, challenge and correct what happened. Agency is distributed across people and rules; it should not be attributed to the most visible actor alone.
Collective modes matter here. Workers can organize to change unsafe procedures; communities can seek public accountability for an automated system. Bandura’s collective agency is not a romantic promise that groups always succeed. It recognizes that some goals require coordination. The Institute’s application is to ask which routes for coordinated action are actually available, who bears the costs of using them and whose voice is excluded.
Why agency cannot be assigned as a personality trait
It is tempting to call someone “agentic” because they are assertive, decisive or optimistic. Those qualities may help in some contexts, but they do not by themselves establish meaningful agency. A quiet person can deliberate carefully and act effectively. An outspoken person can remain powerless in a rigid system. A trait label may hide institutional barriers and encourage blame when people cannot achieve outcomes they never had a fair chance to influence.
The sources used here are conceptual syntheses and reviews across psychology, sociology and work research. They support a multidimensional definition, not a diagnostic profile. Different fields use “agency” with different emphases. Some study intentional action; some study social change; some study motivation or legal rights. The Institute’s synthesis is to keep capacity, available options and effective consequences in one conversation while stating which part a particular study actually measured.
This perspective also leaves room for disagreement. A policy can expand one option while imposing a cost on another. A choice can be informed yet constrained by poverty or care duties. What counts as a valued end may differ among people. Agency research cannot determine those values from a generic survey. It can help make the distribution of choices and power visible so affected people can deliberate about them.
What the evidence does not show
These sources do not establish a single universal agency scale or a fixed amount of agency that belongs to a person across all settings. They do not show that confidence, felt control or the presence of a choice button is sufficient. They also do not show that intention guarantees a desired outcome. The literature spans distinct levels of explanation, from psychological self-regulation to social structure. A claim about one level must not be silently promoted to another.
The Institute’s design questions are proposals, not interventions tested together in Bandura’s, Emirbayer and Mische’s or Deci and colleagues’ reviews. To demonstrate that a particular policy expands agency, researchers would need to examine usable options, actual decision power and effects for the people concerned, including those least well served by the existing system. More training alone is not proof of structural change.
Return to the contested decision
The person who sees the error has knowledge and a purpose. If they can inspect the record, reach a reviewer and change the result, those capacities may become action. If every route is closed, the missing element is not a better feeling about choice. It is a workable path through an institution. That is why agency belongs to both inner capacity and outer conditions.
A serious account asks who can form a goal, what options they can use, what support they may enlist and whether their action can matter. It also asks whether the person endorses the aim and how they experience their role. The terms are connected, but each adds a distinct piece. Keeping them separate makes agency a useful concept for design and research rather than a compliment attached to any final click.
Research references
- Bandura, A. (2001). Social cognitive theory: An agentic perspective. Annual Review of Psychology, 52, 1–26. View source
- Bandura, A. (2006). Toward a psychology of human agency. Perspectives on Psychological Science, 1(2), 164–180. View source
- Emirbayer, M., & Mische, A. (1998). What is agency? American Journal of Sociology, 103(4), 962–1023. View source
- Deci, E. L., Olafsen, A. H., & Ryan, R. M. (2017). Self-determination theory in work organizations: The state of a science. Annual Review of Organizational Psychology and Organizational Behavior, 4, 19–43. View source
- Haggard, P. (2017). Sense of agency in the human brain. Nature Reviews Neuroscience, 18, 196–207. View source