Agency

Sense of agency:
what it means, how to keep it

Sense of agency is the feeling that you are the one doing what you do, and that what follows is your doing. You notice it mostly when it fails: a cursor that lags, a car that brakes by itself, a paragraph a chatbot wrote under your name.

Human agency, on this site, is a capacity: setting the direction of your work and acting on it. Sense of agency is that capacity felt from the inside, and the two can come apart.

What does sense of agency mean in psychology?

In psychology, sense of agency means the experience of controlling your own actions and, through them, the course of events outside you. That is Patrick Haggard’s definition (2017); James Moore (2016) puts it more briefly: “the feeling of control over actions and their consequences”.

• Agency is not ownership. Shaun Gallagher separated the two in 2000: agency is the sense that you are the source of an action, ownership the sense that it is your body that moves. If someone pushes your arm, the arm is yours, the movement is not.
• A feeling is not a judgement. Matthis Synofzik and colleagues (2008, as summarised by Moore) split it into a feeling of agency, the wordless sense of control while you act, and a judgement of agency, the explicit “I did that”.

How does the brain produce the feeling of control?

The brain produces the feeling of control by checking what it predicted against what happened. In the comparator model (Chris Frith and colleagues, 2000, as summarised by Moore, 2016), the motor system predicts the sensory result of its own command; a match feels like “mine”.

The best-known measurement is by Haggard, Clark and Kalogeras (2002). Nine people pressed a key that produced a tone a quarter of a second later. The press was felt 15 milliseconds later and the tone 46 milliseconds earlier than when each occurred alone: action and effect were pulled together in perceived time, an effect called intentional binding.

What this does not show. Nine people, single key presses, and what the effect measures is disputed. Keisuke Suzuki and colleagues (2019) found the same compression with no intentional action, and Gaiqing Kong and colleagues (2024) found no evidence that intention produces it; a 2024 replication of Suzuki’s study by Wiesing and Zimmermann found the opposite, so the question is open. Read binding with caution, not as a direct readout of agency.

The feeling can be wrong. A second account (Daniel Wegner, as summarised by Moore, 2016) holds that the feeling is inferred after the fact, when a thought comes just before an action and fits it. Feeling and action can also come apart in illness: Moore cites patients with schizophrenia who move their own hand and feel they do not control it.

Sense of agency vs locus of control vs self-efficacy: what is the difference?

Sense of agency is an experience during action, locus of control is a general belief about what causes outcomes, and self-efficacy is a belief about what you can do.

The two beliefs. Julian Rotter’s 1966 monograph concerns “generalized expectancies”: whether you expect what happens to you to follow from your own behaviour or from luck, fate and powerful others. For Albert Bandura (1977), expectations of personal efficacy decide whether you start and how long you persist; the first of their four sources is performance accomplishments.

How they connect. Adam Tapal and colleagues (2017) built the Sense of Agency Scale, which asks about the feeling in general, not about one action (236 and 408 participants, in Hebrew). In the second sample, a stronger sense of positive agency went with a less external locus of control (r = -0.35), and scores overlapped only partly with general self-efficacy: related but different, the authors concluded.

What happens to your sense of agency when AI does the work?

When a machine took over more of a task, in the one study that measured it step by step, the sense of agency fell; with AI-written text, people report little ownership.

• Automation. Bruno Berberian and colleagues (2012) had 13 people without flying experience supervise a simulated flight at four levels of automation. The more the system did, the weaker the binding and people’s own ratings of authorship: the longer the interval felt, the lower the rating (mean r = -0.84). Does not show: generative AI.
• Writing with AI. Fiona Draxler and colleagues (2024) ran two studies (30 and 96 people) on postcards written with GPT-3. People felt little ownership of the text; more influence on it meant more. Does not show: sense of agency as such (the results are ownership, declared authorship and a control rating), or effects over time.
• Help that feels like you. Wen Wen, Atsushi Yamashita and Hajime Asama (2015) had 17 students steer a dot while a computer, without telling them, ignored their wrong commands. When the dot answered with a delay of 400 or 700 ms, performance and the reported sense of control both rose, although about 42% of commands on average were never executed. Does not show: any effect at a 100 ms delay, or anything outside a simple laboratory task.

As of 9 November 2026 we found no published study of intentional binding during work with a generative AI assistant. On skill, see cognitive offloading.

How can you practise your sense of agency in five minutes?

You can exercise your sense of agency with five minutes of writing around one small action: write down what you expect, do the action, and compare the result with what you wrote. If you came asking how to take control of your life, this page starts smaller: one action you can check today. We call it “call it, do it, check it”.

1. Pick one action for the next hour that depends mostly on you. A paragraph, a call, a postponed decision.
2. Call it. Write one line: “I will do X, and I expect Y.” If AI helps, add which decision stays yours.
3. Do it. If you use AI, change the result in at least one way of your own: cut, reorder, reject, rewrite.
4. Check it. Write one line: what happened, and which part came from what you did. Mark it match, partly or miss.
5. After a miss, write one thing you would do differently.

A match (invented example). Action: tell the client about the delay. I expect: a new date, sent by 3 pm. AI: first draft. I decide: date and tone. What happened: sent at 2:40. My part: the date, one cut.

A miss. Action: one-page summary. I expect: three points in 30 minutes. What happened: I pasted the AI summary; it left out the budget risk. My part: I pressed send. Next time: write my three points first.

Why it should work. In the comparator account, the feeling of agency during a movement comes from a prediction that matches its result. The log borrows that idea by analogy: a written prediction gives your judgement something definite to compare. A record of things you called and then did resembles Bandura’s first source of efficacy, and recording misses keeps the feeling tied to what you really caused.

The limit. No study we know of has tested this log as an intervention. It is a way to pay attention, not a tested method. Related: five questions before you ask AI, acting under uncertainty.

What does this not promise?

This page does not promise control over your life. Neither does the research: none of the studies here shows that a stronger sense of agency changes income, career or happiness.

• A feeling of control is not control. In the steering study the feeling rose while control fell.
• Some things are not up to you. On sorting what is yours from what is not, see the dichotomy of control.
• This is not treatment. A lasting sense that your actions are not your own, or helplessness that gets in the way of daily life, is a reason to talk to a doctor or a licensed mental health professional.

Where does the question come from?

The question of what is up to a human being is far older than the term: Hermetic texts composed in Greek in Roman Egypt already ask it.

Corpus Hermeticum IV. In the treatise of the mixing bowl, Hermes tells Tat that reason was given to everyone, but mind (nous) was set out as a prize for those who choose it. Section 8, in G.R.S. Mead’s 1906 translation: “that which is dependent on ourselves, let this press onward and have no delay”.

The Asclepius and Pico. The Asclepius, which survives complete only in Latin, says in chapter 6: magnum miraculum est homo, “a great miracle is man”, a being earthly in one part and divine in the other. Giovanni Pico della Mirandola (1463-1494) quoted it in the speech he wrote in 1486, often read as a manifesto of human freedom and, by Brian Copenhaver (2019), as a text about ascent toward angelic existence.

How far the bridge goes. These texts are the source of the question, not evidence that a practice works. They speak about choice, not a measurable feeling, and this page claims no line of descent from them to modern psychology. They share a question: which part of what happens is yours? See operational Hermeticism.

Key facts

Definition	The experience of controlling your own actions and, through them, events in the world (Haggard 2017)
Not the same as	Ownership (Gallagher 2000), locus of control (Rotter 1966), self-efficacy (Bandura 1977)
Implicit measure	Intentional binding (Haggard and colleagues 2002; 9 people); what it measures is disputed (Suzuki 2019; Kong 2024)
Under automation	Weaker with each level of automation (Berberian and colleagues 2012; 13 people)
Old texts	Corpus Hermeticum IV; Asclepius 6; Pico’s Oration (written 1486)
Checked	9 November 2026

Fair questions

What is a sense of agency in simple terms?

It is the feeling that you are the one doing what you do and causing what follows. Psychologists define it as the experience of controlling your own actions and, through them, events in the outside world (Haggard 2017).

What is the difference between sense of agency and locus of control?

Sense of agency is an experience tied to a particular action: did I do that? Locus of control, the construct Julian Rotter set out in a 1966 monograph, is a general belief about whether what happens to you follows from your own behaviour or from luck, fate and powerful others.

Is sense of agency the same as self-efficacy?

No. Self-efficacy, Albert Bandura’s term from 1977, is a belief about whether you can carry out a behaviour, held before you act. Sense of agency is the experience of control during and after the action. The two are related but distinct: in a 2017 questionnaire study their scores overlapped only partly.

Does using AI weaken your sense of agency?

No study on this page measured that directly for generative AI. In a 2012 flight-supervision study with 13 people, the sense of agency fell as automation took over more of the task. In two 2024 studies of AI-written postcards people felt little ownership of the text unless they had shaped it, which is a related feeling, not the same one. Quiet assistance can also raise the feeling while you control less, so the feeling alone is not a reliable gauge.

How do I take control of my life?

Nobody controls a whole life, and no study on this page promises that. A workable start is small: pick one action that depends mostly on you, write down what you expect, do it, and compare the result with what you wrote. If helplessness gets in the way of daily life, talk to a doctor or a licensed mental health professional.

Sources & further reading

• Haggard, P. (2017). Sense of agency in the human brain. Nature Reviews Neuroscience 18, 196-207. doi:10.1038/nrn.2017.14
• Moore, J. W. (2016). What is the sense of agency and why does it matter? Frontiers in Psychology 7, 1272. doi:10.3389/fpsyg.2016.01272
• Gallagher, S. (2000). Philosophical conceptions of the self. Trends in Cognitive Sciences 4(1), 14-21. doi:10.1016/S1364-6613(99)01417-5
• Haggard, P., Clark, S. and Kalogeras, J. (2002). Voluntary action and conscious awareness. Nature Neuroscience 5(4), 382-385. doi:10.1038/nn827
• Suzuki, K. and three co-authors (2019). Intentional binding without intentional action. Psychological Science 30(6), 842-853. doi:10.1177/0956797619842191. Opposite result: Wiesing, M. and Zimmermann, E. (2024). Consciousness and Cognition 119, 103665. doi:10.1016/j.concog.2024.103665
• Kong, G. and four co-authors (2024). No evidence in favor of the existence of “intentional” binding. JEP: Human Perception and Performance 50(6), 626-635. doi:10.1037/xhp0001204
• Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied 80(1), 1-28. doi:10.1037/h0092976
• Bandura, A. (1977). Self-efficacy: toward a unifying theory of behavioral change. Psychological Review 84(2), 191-215. doi:10.1037/0033-295X.84.2.191
• Tapal, A. and three co-authors (2017). The Sense of Agency Scale. Frontiers in Psychology 8, 1552. doi:10.3389/fpsyg.2017.01552
• Berberian, B. and three co-authors (2012). Automation technology and sense of control. PLoS ONE 7(3), e34075. doi:10.1371/journal.pone.0034075
• Wen, W., Yamashita, A. and Asama, H. (2015). The sense of agency during continuous action. PLoS ONE 10(4), e0125226. doi:10.1371/journal.pone.0125226
• Draxler, F. and six co-authors (2024). The AI Ghostwriter Effect. ACM Transactions on Computer-Human Interaction 31(2). doi:10.1145/3637875
• Copenhaver, B. P. (2019). Magic and the Dignity of Man. Harvard University Press. Library of Congress
• Mead, G.R.S. (1906). Thrice-Greatest Hermes, vol. II, p. 89 (Corpus Hermeticum IV, §8; Mead’s IV (V), The Cup or Monad). Internet Archive