Intention and Belief

The Problem: An Agent Who Intends and Predicts

An agent announces, “I intend to finish the manuscript tonight.” The announcement expresses a resolution about what to do. It may also invite the prediction that a finished manuscript will exist by midnight. The relation between those two attitudes is the subject of this investigation. Does the intention itself entail the belief that the manuscript will be finished? Can someone intend to finish while reasonably believing that illness, interruption, or poor concentration may prevent completion? Could an intention itself constitute the relevant belief, or do intention and belief exercise different kinds of rational authority?

Three features make the problem difficult. First, intention concerns prospective conduct, so its content is frequently expressed in the same future-tense form used for prediction. Second, intending appears to settle deliberation and provide a basis for planning; an intention wholly disconnected from any expectation that its agent will do something would seem incapable of organizing later conduct. Third, human agents often intend difficult actions whose outcomes remain uncertain. A theory that builds confidence into every intention must explain these imperfectly controlled cases rather than dismiss them as failures of sincerity.

The inquiry therefore concerns the kind and strength of an epistemic commitment generated by practical commitment. Its central alternatives are (a) no necessary belief relation; (b) a belief relation required by rationality but not by the essence of intention; (c) belief as a necessary constituent of intention; and (d) intention itself as a practically grounded form of belief. These alternatives differ in their metaphysics of attitudes and in the norms they use to evaluate a person who intends to act under risk (Bratman, 1987, 2009; Holton, 2008; Marušić & Schwenkler, 2018).

A satisfactory answer must explain an ordinary successful intention, an abandoned intention, an intention that governs a series of preparations, and an intention whose execution is uncertain from the outset. It must also identify whether the belief at issue concerns an outcome, an attempted action, an action now in progress, or the agent’s present commitment. Each proposition has different evidence and success conditions.

Definition and Scope of the Two Attitudes

Intention, in the present study, is a practical attitude by which an agent settles upon some action or end in a manner that ordinarily guides later deliberation and performance. This working characterization includes prospective intentions and the intention with which one acts, while leaving open the competing metaphysical accounts of the attitude. It distinguishes intending from mere wishing: a wish can persist without any adopted plan, whereas an intention normally organizes what the agent does next, subject to revision when circumstances change.

Belief is a truth-evaluable cognitive attitude that represents a proposition as being the case. To believe that one will mail a letter tomorrow is to take a proposition concerning a future event as true. The proposition can be false even if the agent has perfectly good evidence. To intend to mail the letter is to undertake an action-guiding commitment whose success is assessed partly by whether the letter is mailed and by the circumstances of the attempt. Practical reasons can justify the intention while quite different evidence justifies the prediction (Davidson, 1978/2001; Bratman, 2009).

The disputed entailment can be represented schematically as I(A) → B(F(A)), where I(A) says that an agent intends to perform action A and B(F(A)) says that the agent believes that she will perform A. The formula represents a philosophical hypothesis. It is not an established logical law. Some accounts require belief that one will try; some require belief that one is currently acting; some require high confidence rather than full belief; and some replace the entailment with a rational coherence requirement. Substituting these different propositions for F(A) changes the theory.

Success must also be disambiguated. Someone can intentionally begin building a house without finishing it, attempt a difficult rescue without achieving the rescue, or intend to send a message and discover that the network has failed. The intention may govern a genuine action while the intended outcome remains unrealized. The distinction between action descriptions and outcomes is indispensable for evaluating the argument from intention to belief (Anscombe, 1957/2000, §§23–26; Holton, 2008).

A Historical Beginning: Aristotle on Choice and Opinion

The classical background is Aristotle’s analysis of prohairesis, conventionally translated as choice or decision after deliberation. In Nicomachean Ethics III.2–3, Aristotle distinguishes choice from appetite, wish, and doxa, or opinion. He argues that opinions may concern things beyond our agency, including the impossible and the eternal, whereas deliberative choice concerns actions within our power. Opinion is assessed for truth and falsity; choice is assessed by the orientation of action toward what is taken to be good (Aristotle, Nicomachean Ethics 1111b–1113a).

This is an important antecedent of the modern intention–belief problem. The distinction is functional and normative: the agent can make an excellent prediction and an objectionable choice, or make an admirable choice while holding inaccurate views about circumstances. Aristotle’s distinction does not already answer whether choosing something includes a belief about one’s future conduct. Its immediate point is that choosing and opining are not interchangeable simply because both can issue from reasoning.

Aristotle also describes deliberation as inquiry into what can be done and by what means. Choice fixes a practical result at the end of that inquiry. It is therefore situated at a boundary between grasping circumstances and undertaking conduct. The practical conclusion depends upon beliefs about available means, impediments, and causal connections. This dependence is not the same as saying that the conclusion itself is a prediction of successful performance.

A conceptual lesson follows. The fact that rational choice relies on beliefs does not demonstrate that choice is a belief. A second and more demanding argument is required to establish the identity or entailment thesis. This argumentative gap survives in later theories that derive intentions from beliefs and desires, or interpret means–end coherence as a disguised rule of theoretical inference. Aristotle offers the initial distinction; the modern debate asks how the two rational orientations are nevertheless integrated.

Medieval Articulation: Intention, End, and the Will

Thomas Aquinas provides another historical model in Summa Theologiae I–II, question 12. He classifies intention principally as an act of the will ordered toward an end, while recognizing that reason specifies an intelligible ordering of means to ends. In article 1, intention expresses tending toward an end. In article 4, the willing of means and the intention of an end can form one movement considered as purposively ordered. This vocabulary does not reproduce contemporary belief psychology, but it identifies a practical structure irreducible to a bare proposition’s truth (Aquinas, Summa Theologiae I–II, q.12, aa.1, 4).

Aquinas’s account clarifies why one can intend an end before determining every means required for achieving it. The agent’s orientation can be settled while the execution plan remains partially specified. It also shows why deliberation can continue after commitment: there remains a problem of how to attain the already intended end. This distinction anticipates the importance of partial plans in modern action theory without implying direct historical identity between the two accounts.

The medieval distinction between willing an end and intellectually grasping a means creates a productive question for belief theories. If the will is directed to an end through apprehended considerations, which cognitive states must the agent possess? Beliefs about the means, about the end’s attainability, and about one’s present decision belong to different categories. A general claim that intention entails belief is too indeterminate until it specifies which of them is intended.

The historical significance of Aquinas lies in his explicit separation and coordination of cognitive and volitional powers. One need not adopt that faculty psychology to preserve the analytical problem it addresses: a practical commitment has representational conditions, but representation alone does not explain the directedness of committed agency.

Anscombe: Prediction, Action, and Practical Knowledge

Elizabeth Anscombe’s Intention (first published in 1957; second edition 1963) transformed the debate by refusing to assume that intention could be isolated as a private psychological episode. She begins with three linked uses: an expression of future intention, an intentional action, and an intention with which one acts. The problem is to articulate their unity without reducing every case to an introspected decision (Anscombe, 1957/2000, §§1–4).

Anscombe identifies intentional action through a distinctive application of the question “Why?” An agent can answer by giving a reason that displays what she is doing under a description significant to her practical undertaking. Her familiar extended example of pumping water illustrates that the same bodily sequence can be described as moving an arm, pumping, replenishing a supply, or bringing about further consequences. The descriptions under which a performance is intentional do not automatically include every result the agent foresees (Anscombe, §§5–8, 23–26).

In §§28–33, the issue becomes knowledge. An agent can often say what she is doing without inspecting herself or inferring the answer from observable bodily evidence. The notion of practical knowledge concerns action as intentionally directed and understood by the agent who carries it out. It poses a difficult question: how can a claim to know what one is doing remain knowledge if the action fails, or if the physical process takes an unexpected course?

Anscombe’s shopping-list comparison distinguishes the role of a list guiding purchases from that of a detective’s list reporting them. A discrepancy between the two calls for different forms of correction. If the shopper buys the wrong item, correcting the performance answers to the shopping list; if the detective records the wrong item, correcting the record answers to what happened (Anscombe, §32). This distinction is a central source for later discussions of directions of fit. It supports a contrast in standards of correctness while leaving substantial interpretive questions about the structure of practical knowledge.

The strongest conclusion justified by the text is that intending and doing intentionally involve a practical mode of knowing or determining action that cannot simply be modeled on observational prediction. Whether Anscombe thinks all intention includes an ordinary belief that an intended outcome will happen is disputed. Her analysis should not be flattened into either a categorical identity between intention and belief or a denial that intending has any epistemic significance.

Davidson: Reasons, Causes, and Pure Intending

Donald Davidson’s “Actions, Reasons, and Causes” (1963) defends the thesis that a primary reason can explain an intentional action causally as well as rationally. A favorable attitude toward an end, together with a belief about how an action would promote it, can rationalize the deed; the rationalizing state must also stand in an appropriate causal relation to what occurs. The presence of a justificatory reason does not by itself establish that it was the reason for which the agent acted (Davidson, 1963/2001).

The initial structure makes beliefs indispensable as explanatory background: the agent must represent a way of obtaining or promoting what is favored. But it leaves open whether the intention is itself a belief. The belief involved in a primary reason can concern what would promote an end, whereas the belief postulated by the intention–belief entailment thesis concerns what the agent herself will do. Conflating these propositions would make the argument circular.

In “Intending” (1978), Davidson confronts prospective intending that exists before any performance has begun. An agent may resolve to undertake work later and yet undertake no preparatory movement now. Such pure intention cannot be identified with an already performed intentional action. Davidson’s subsequent account of unconditional practical judgment attempts to show how a practical evaluation can settle upon an action in a way that differs from an isolated desire (Davidson, 1978/2001).

This account generates two enduring questions. One concerns whether a person can rationally choose between equally acceptable options without judging the chosen one better than the alternatives. The other concerns the relationship between the reasons for intending an action and evidence that the action will occur. The latter is especially pressing because poor evidence of likely success need not eliminate a sufficiently strong reason to try. Davidson’s intervention thus makes visible a gap between the rationalizing grounds of agency and ordinary grounds of prediction.

Bratman: Plans, Stability, and the Independence of Intention

Michael Bratman’s Intention, Plans, and Practical Reason (1987) treats intentions as elements of plans that coordinate temporally extended conduct. Agents have limited deliberative resources and cannot continuously reopen every settled question. Intentions therefore support stability, organize later subplans, filter incompatible options, and enable coordination with others. These functions explain why an intention has a distinctive role even when the agent possesses relevant beliefs and desires (Bratman, 1987; 1999).

Planning theory introduces requirements of internal consistency and means–end coherence. If an agent intends to attend an appointment and knows that travel is necessary, she has a practical problem to solve about travel. Her failure to adopt or arrange means, absent an acceptable explanation, renders the practical organization of her attitudes defective. A belief about a route may guide the plan, but a belief alone ordinarily does not settle the action to be taken.

Bratman’s challenge to the simple view of intention is instructive. Someone may reasonably attempt two mutually exclusive goals, knowing that only one can succeed, when doing so maximizes the prospect of obtaining at least one important outcome. It would be implausible to require two jointly inconsistent full commitments to successful achievement solely because either successful attempt would count as intentional. A distinction between intending to try and intending to achieve may preserve rational consistency (Bratman, 1987, chap. 8; Setiya, 2022).

In his later criticism of cognitivism, Bratman argues that the practical norms governing intentions need not derive from theoretical norms governing beliefs. Their point can be explained by the organizational functions of plans themselves. The fact that intentions constrain what agents may coherently believe and pursue does not establish that they are beliefs in disguise. A planning theory can therefore acknowledge epistemic implications while maintaining a distinct account of practical commitment (Bratman, 2009).

An important qualification is that the intention to try is not always a complete substitute for an intention to succeed. There are circumstances, including some promises, in which the end itself constitutes the undertaking. An adequate planning view must explain when agents accept responsibility for the result, when they undertake only an effort, and when the content of a plan contains both.

Velleman: Intention as Self-Fulfilling Expectation

J. David Velleman’s Practical Reflection (originally 1989) develops a form of cognitivism in which reflection about one’s own future conduct can help constitute that conduct. The agent’s expectation is not a passive forecast; it can contribute to its own realization through the practical motivation associated with making one’s conduct intelligible to oneself. This proposal challenges a picture in which desire, belief, decision, and bodily movement follow one another as externally connected events (Velleman, 1989/2007).

The self-fulfilling structure gives a possible answer to the problem of how belief can have an action-guiding role. If I expect to act in a particular way partly because I am deliberating about what to do, that expectation can stabilize my subsequent activity. The cognitive state belongs to a system in which its correctness is affected by its role in guiding action. Its warrant may consequently have a form different from the warrant of an observer who predicts my conduct from my past behavior.

But the theory faces a discrimination problem. Many beliefs about one’s future conduct fail to function as intentions. I might expect to forget a birthday, make a mistake, or react nervously, while firmly intending to prevent the expected event. If self-fulfilling expectation explains intention, an additional principle must specify which expectations count as practical commitments and why others remain predictions. Bratman’s criticism of excessive motivational reach addresses this difficulty (Bratman, 1991, as discussed in Setiya, 2022).

Velleman’s 1992 discussion of agency sharpens a related issue. A sequence of desire, instrumental belief, and bodily causation can omit an intelligible account of what makes the agent the author of the action. A theory that assigns an active role to reflective understanding may help restore that dimension, although it must still distinguish legitimate control from merely expected conformity (Velleman, 1992). The philosophical challenge is to explain agency through reflection without giving every self-prediction authority over action.

The Belief Entailment Thesis and Its Forms

The general thesis that intending to do A entails believing that one will do A appears plausible in routine cases. When I deliberately resolve to open a door, my ability, proximity, and ordinary control make failure unlikely. I may naturally answer “I will open it” rather than “I will try to open it.” The explanation may be pragmatic: in expressing the commitment I normally convey confidence as well. It may instead be constitutive: intending requires belief about one’s future action. Or it may be rational: rational intention generally warrants an expectation without being identical to it.

These readings yield distinct predictions. The constitutive entailment thesis excludes any genuine intention accompanied by a contrary unqualified belief. The rational connection thesis allows psychologically possible but normatively defective combinations. The pragmatic thesis allows a speaker to intend while withholding belief, even if ordinary conversation usually encourages a stronger assertion. A more flexible epistemic thesis says that intending normally changes one’s degree of confidence, although not always sufficiently to produce all-out belief (Holton, 2008; Setiya, 2022).

A precise argument for entailment needs two independently supported premises. First, an intention commits the agent to a certain practical description of what she will do. Second, such a commitment must carry the same truth-directed endorsement as belief in the corresponding future proposition. The first is plausible in a wide range of cases. The second is the disputed bridge principle. Showing that commitments rationally govern action does not yet establish that they satisfy the epistemic constitutive conditions of believing.

A counterexample must be evaluated with equal care. If someone says, “I intend to climb the mountain, although I doubt that I will reach the summit,” the doubt could concern a remote outcome while the intention concerns an earnest effort. But it could also be a genuine intention to reach the summit despite substantial risk. Rather than automatically translating the intention into an intention to try, the philosopher must establish what the agent settles, how she plans, and which forms of failure she recognizes as failures of the project.

Strong Cognitivism and Practical Belief

Berislav Marušić and John Schwenkler defend strong cognitivism in “Intending Is Believing” (2018). Their proposal identifies an intention with a belief about what the agent will do when the belief is held on the basis of practical reasoning. The manner of grounding is essential. A predictive belief based on statistical evidence is not automatically an intention; a practically settled belief about what one will do has a different rational origin and role (Marušić & Schwenkler, 2018).

The theory offers conceptual economy. It explains the assertoric appearance of future-directed intention and derives a strong connection between intention and the epistemic standpoint from which one anticipates one’s own conduct. It also promises a unified account of avowal: the person who decides what to do can apparently know and announce what she will do by exercising agency, without waiting for independent evidence of her own future behavior.

Its hardest question concerns the relationship between practical and theoretical rationality. If the same state is belief, how can forming it on practical grounds avoid epistemic irresponsibility when ordinary evidence suggests failure? One response emphasizes a change in the way agents determine what will happen: while the action remains up to them, practical reasoning can settle its future truth in a way that observation cannot. The justification then depends on the agent’s effective capacity, not merely on favorable probability estimates.

Strong cognitivism still needs a criterion that excludes arbitrary self-assurance. A person can become certain that she will win a competition solely by wanting to win; that confidence need not express competent practical reasoning. The theory must specify which commitments constitute agency and which confidence claims are wishful thinking. It must also account for agents who genuinely organize their lives around difficult projects while refusing to claim certainty about success.

These are substantive disputes about the identity and normativity of mental states. The strong cognitivist does not simply assert that all future beliefs are intentions; the proposal identifies a special practical grounding. The critic must therefore show why even the practically grounded subclass retains features irreducible to belief, or why some intentions lack the belief content that the account makes essential.

Evidence, Entitlement, and the Threat of Wishful Thinking

The epistemic objection to cognitivism is often stated as a question of warrant. An agent deciding to perform A may have good practical reasons to perform A but little independent evidence that she will succeed. A spectator with the same statistical information might rationally predict failure. If the agent acquires an unqualified belief in success merely by deciding, why is this belief epistemically justified?

H. P. Grice’s “Intention and Uncertainty” (1971) forcefully raises this problem. One route is to infer what one will do from knowledge of one’s intention and independent beliefs about one’s ability and circumstances. On this reconstruction, intention and belief are distinct, and belief about success requires further epistemic support (Grice, 1971). Such an inference explains many ordinary cases but leaves open the practical force of a decision that itself changes the evidence about future performance.

An alternative holds that practical belief can be warranted through its connection to the agent’s capacity to make it true. The relevant fact is not simply that a belief could cause its own confirmation. Someone might manipulate herself into accepting an arbitrary forecast. The stronger claim is that competent agency supplies a non-observational route to knowledge: an intention organizes ability, attention, and means under the agent’s control. Whether that route is sufficiently secure in a particular case depends on what the action requires (Setiya, 2016; Marušić, 2015).

Sarah K. Paul’s critique of practical-knowledge accounts identifies the danger of treating an act of commitment as if it automatically created its own epistemic warrant. It presses the question of how a practically effective disposition differs from wishful thinking (Paul, 2009; Setiya, 2016). The challenge is productive because it demands a theory of success-sensitive warrant rather than a verbal reclassification of unsupported confidence.

The resulting distinction is between authorization to undertake an action and justification for expecting its completion. A good practical reason can authorize attempting a difficult rescue without generating high confidence in the rescue’s success. A theory that denies this difference converts courage into irrational prediction. Conversely, a person’s evident competence and successful control can justify confidence in a nearby action even where very little external prediction precedes it.

Practical Knowledge: What an Agent Knows While Acting

The thesis of practical knowledge should be examined separately from a prospective belief that an action will be completed. When an agent is intentionally writing a sentence, she can often say what she is doing immediately. This knowledge has a practical source: she is directing the process under the relevant description. It is not ordinarily inferred from watching her fingers or checking the screen. But knowing that one is now writing a sentence is distinct from knowing that the sentence will be finished or accepted by a journal.

Anscombe’s account is the classic source for the connection, and Kieran Setiya’s work develops a contemporary explanation. In Practical Knowledge: Selected Essays (2016), Setiya argues that knowledge of what one is doing can arise from a practical capacity involving know-how rather than from antecedent observational evidence. The argument ties the epistemic relation to competent intentional performance; it does not make every imagined future success into knowledge (Setiya, 2016, chaps. 1, 4).

This framework also raises the problem of error. An agent may believe she is opening a valve that is jammed, signing a document with an empty pen, or crossing a boundary she has not yet reached. Sometimes she knows she is trying, although the intended result has failed. Sometimes she knows an activity is underway in an imperfective sense that does not require its eventual completion. Sometimes her self-description is simply mistaken. A theory that cannot allow all three possibilities will misdescribe practical fallibility.

The relevant grammar distinguishes “I am writing,” “I am writing a book,” and “I have written a book.” Ongoing activity can be genuinely intentional while the larger accomplishment is incomplete. This matters because the most attractive practical-knowledge thesis may concern action under a description appropriate to its unfolding, whereas the disputed entailment often concerns a future perfective outcome. A philosophical inference from one to the other requires additional premises about ability and completion.

Practical knowledge consequently suggests a graded family of epistemic relations. There is immediate knowledge of adopted commitment, non-observational knowledge of ongoing competent action, justified expectation about likely continuation, and empirical knowledge of finished results. These relations can support one another, but they should not be treated as one undifferentiated belief.

Uncertainty and Holton’s Partial-Belief Challenge

Richard Holton’s “Partial Belief, Partial Intention” (2008) attacks the assumption that the issue must be framed solely in terms of all-or-nothing belief. People routinely act with intermediate confidence. An agent might intend to meet a friend at an outdoor event despite a substantial possibility of disruption, or intend to complete a demanding task without assigning near certainty to completion. The strength of expectation varies while the commitment continues to structure practical deliberation.

Holton argues that the important connections between intention and belief are normative rather than essential. He also develops a concept of partial intention that interacts with partial belief. His approach moves the problem from a crude test of whether all intentions carry categorical belief to an investigation of how attitudes are integrated under uncertainty (Holton, 2008).

The significance is not merely psychological. If an intention is stable while an agent’s confidence fluctuates, identifying the intention with a fixed level of credence becomes difficult. Confidence may fall after new evidence without immediately reversing the practical decision. Conversely, confidence in a likely event may increase without generating any corresponding intention. This independence suggests that an agent’s commitments and forecasts occupy distinct dimensions even when they rationally influence one another.

A graded theory must avoid a second oversimplification. It would be equally mistaken to require that intending always raises the credence of success by a fixed amount. Some intentions express commitment despite unchanged objective risk, especially where other agents or uncontrollable circumstances determine the outcome. Forming the intention may improve the prospect of success by ensuring effort and preparation, but it cannot guarantee that the external world will cooperate.

The useful result is a model in which confidence, strength of commitment, and degree of control are separate variables. Rational relations among them are context-sensitive. This permits stable intentions under uncertainty without treating them as epistemically inert: a sincere intention normally matters to how an agent assesses what she is likely to do, but its evidential force depends on the mechanism by which the action would be realized.

The Difference Between Intending to Act and Intending to Try

An intention to try is directed at undertaking an attempt; an intention to accomplish is directed at bringing about an end. In many routine activities the two overlap. When a person intends to lift an ordinary cup, the intention normally governs the lifting and supports confident expectation that the cup will rise. In difficult undertakings the distinction becomes significant. One can genuinely try to rescue a stranded climber while knowing the rescue may fail.

A theory of intention and belief should ask which proposition the agent is committed to and what would count as its satisfaction. If the intended object is trying, the belief that one will attempt an action is compatible with uncertainty about success. If the intended object is successful rescue, the agent undertakes a more demanding aim and evaluates the outcome by that end. An unqualified belief about achieving the end need not follow merely from the agent’s commitment to pursuing it.

Bratman’s cases of jointly incompatible goals bring the distinction into focus. One may intentionally attempt two routes while knowing they cannot both succeed, provided the projects are coordinated in a way that makes the combined attempt rational. The inference from intentionally performing a route-specific attempt to possessing two contradictory beliefs about ultimate success should therefore be rejected (Bratman, 1987, chap. 8).

Yet the “try” solution has limits. A person may promise to deliver crucial medication, with the understanding that the end rather than mere effort is the relevant undertaking. Replacing the promise with a promise to try can change its normative content. Marušić’s examination of promising and resolving against evidence makes this point central: genuine commitments often concern performing the action, even when there is evidence of possible failure (Marušić, 2015).

There is thus no universal rule that converts every risky intention into an intention merely to try. The correct analysis respects the agent’s adopted practical object, the action’s internal and external conditions, and the circumstances that would excuse rather than erase the failure. The belief appropriate to the case must be evaluated only after this intentional content is determined.

Means–End Coherence and Two Forms of Rationality

Suppose an agent intends to attend a conference, believes taking a train is necessary, and has neither booked the train nor adopted any other plan capable of securing travel. The difficulty is immediately recognizable as a defect in practical organization. The crucial question is whether that defect can be reduced to a violation of theoretical rationality governing beliefs.

A cognitivist explanation begins with the thesis that intending to attend entails believing one will attend. If the agent also believes that attending requires travel, ordinary deductive coherence seems to put pressure on believing that suitable travel will occur. This promises to derive means–end coherence from consistency or closure in belief. But a gap remains: believing that one will book a ticket is not the same as booking one or intending to book one. The theorist needs a bridge from epistemic coherence to action-guiding commitment (Bratman, 2009).

Bratman’s planning account instead treats means–end coherence as a practical norm that enables plans to guide conduct effectively. An intention supplies a problem; beliefs about the world locate possible solutions; and further intentions can complete the plan. The coherence requirement is not just a demand to accept a logically implied proposition. It is a demand to organize conduct around feasible means in a way that preserves the purposes for which the plan exists.

The two explanations may converge in easy cases while diverging under uncertainty. A person can have a consistent set of predictions and still be practically idle, while another may pursue a worthwhile project under substantial predictive uncertainty. The latter can be practically coherent without possessing categorical beliefs that all intended results will happen. The former can be epistemically coherent while remaining defective in agency.

The central finding is that theoretical rationality and practical rationality constrain the same agent through different immediate tasks. The former assesses what should be accepted as true; the latter assesses what should be chosen, maintained, revised, and done. Their interaction is systematic because action depends on information, but reducing one to the other requires an argument that respects these differing success conditions.

First-Person Avowal, Prediction, and the Future Tense

A person ordinarily knows whether she intends to send a letter, although she may be uncertain whether she will actually send it. She can say, “I am going to send the letter,” and conversational circumstances may treat this as an expression of intention. An observer may use the same words to predict her behavior. The grammatical similarity invites the belief-entailment thesis, but grammar alone cannot establish the metaphysical identity of the attitudes.

The authority of first-person avowal has at least two sources. One is the person’s special access to an already formed intention. The other is the authority involved in making or reaffirming a decision in speech. Saying that one will do something can help settle the practical matter when it is still open to decision. Such authority concerns the speaker’s commitment and prospective control, and differs from the authority of a detached forecast.

A confident avowal may be misleading when the speaker ignores predictable obstacles. A cautious avowal may be perfectly sincere when the project is difficult. The phrase “I intend to do it, though I may fail” is coherent in many contexts; its significance depends on whether the speaker merely doubts success, expects failure, or believes an attempt itself is impossible. Grice’s discussion of uncertainty demonstrates that ordinary expressions of intention can create pragmatic expectations that require careful philosophical interpretation (Grice, 1971).

It is also possible to predict a future event without intending it. A person may believe she will stumble, lose patience, or forget something, and nevertheless plan to prevent it. This is a decisive constraint on identity accounts. Even if some practically grounded beliefs count as intentions, not every prediction of one’s own conduct does. The proper theory therefore needs a constitutive role for practical direction or grounding that an ordinary forecast lacks.

The Evidential Problem of Promises and Resolutions

Promises and resolutions expose an apparent tension between practical commitment and unfavorable evidence. A person resolving to resist a familiar temptation may have excellent evidence that similar resolutions have failed. A categorical prediction of success could be overconfident. Yet announcing an intention while insisting that failure is inevitable may undermine the resolution’s sincerity or practical stability.

In Evidence and Agency (2015), Marušić examines this tension and defends the possibility of rational commitment despite adverse evidence. His account treats the deliberative standpoint as one from which agents can settle what is up to them, rather than only calculating how they have behaved previously. The solution does not require that the future be under unlimited control; it calls attention to the way an agent’s present decision can matter to what eventually occurs (Marušić, 2015).

The philosophical strength of this response is that it does justice to commitment as a transformative factor. A decision to exercise every day changes habits, schedules, and the evidential basis for future predictions if it is effectively integrated into conduct. Its weakness, unless carefully qualified, is the possibility of turning intention into an exemption from ordinary evidence. Repeated inability, severe constraints, or dependence on uncooperative others should sometimes reduce rational confidence even in the presence of sincere commitment.

A normatively serious account must distinguish a promise’s content, the promisor’s resolve, her grounds for expecting compliance, and her responsibility when she fails. These considerations overlap but are not equivalent. The rationality of making a difficult promise cannot be established solely by demonstrating that a successful outcome is more likely than not. Equally, confidence that failure is possible does not dissolve an agent’s responsibility to make a serious effort.

Causal Guidance, Deviance, and the Role of Belief

The intention–belief problem also intersects with the causal explanation of action. Suppose an agent wants to signal for help and believes that pressing a button will activate a beacon. She intends to press the button, but anxiety causes an involuntary spasm that happens to strike it. The button is pressed, yet the relevant bodily motion did not occur in the way her intention was supposed to guide it. The agent’s belief about the means can be correct, and the intended result can occur, without establishing that she intentionally performed the act under the relevant description.

Davidson’s causal theory and later discussions of causal deviance make this distinction indispensable. It is not enough for a reason or intention to be somewhere in the causal history of a result. The relation must sustain control or guidance of the sort relevant to agency (Davidson, 1963/2001; Velleman, 1992). A theory that identifies intentions with beliefs still owes an explanation of this directed causal or constitutive relation.

A belief about success can contribute to confidence, persistence, and the selection of means. But confidence itself can be causally counterproductive when it discourages contingency planning. Conversely, accurate uncertainty can improve performance by directing attention to failure modes. Thus the practical value of believing that one will act is conditional on the feedback relation between expectation and execution.

The richer explanatory unit is a temporally extended process. An intention selects an end; beliefs locate opportunities and obstacles; subintentions organize means; ongoing perception corrects execution; and performance succeeds or fails under the adopted description. The relation between belief and intention changes across these phases. An account restricted to the mental state immediately preceding bodily movement cannot capture the full practical structure.

Action Under Risk: Three Diagnostic Cases

Consider a reliable agent who intends to turn on a lamp. The act is ordinary, the switch works, and there are no evident obstacles. It is rational to believe that the lamp will be switched on. The intention is one reason for that belief, since the agent is competent and the task is under her control. Nevertheless, the mere truth that intention and belief coexist does not settle whether the belief constitutes the intention. Their relationship may be one of normally justified expectation.

Now consider an engineer attempting to restore a damaged system with a one-in-three chance of success. She intends to restore the system and has a concrete intervention plan. She also believes that failure is more probable than success. This does not automatically show that her practical state is incoherent. It becomes necessary to ask whether the intended object is restoration, the attempt, or some particular step, and whether the undertaking gives her sufficient reason to proceed despite limited prospects. If success is beyond her control, confidence need not match the seriousness of commitment.

Finally, consider a person who predicts that she will miss an appointment because she habitually procrastinates. She sincerely intends to attend and begins arranging transport and reminders. Her predictive belief is supported by past evidence, while her present intention attempts to change the very process from which that evidence was drawn. Practical and theoretical perspectives concern one developing course of events; the self-prediction may rationally change as the plan acquires effective means.

The cases distinguish four variables: content of commitment, expected probability, degree of agent control, and stage of execution. They also show why no single unqualified proposition about eventual success captures every epistemic consequence of intending. A theory should predict the interaction of these variables rather than simply stipulate that all legitimate intentions entail certainty.

Contemporary Debate: What Can Each Theory Explain?

Noncognitivist planning theory explains why intention settles deliberation, preserves projects across time, and coordinates means without requiring categorical success belief. Its principal explanatory task is to account for why rational intentions often give agents unusually authoritative expectations about what they will do. Such expectations are not arbitrary external additions: they arise from the role of the commitment in controlling future performance (Bratman, 1987, 2009).

Strong cognitivism explains the proximity between deciding, avowing, and believing that one will act. It puts practical reasoning inside the account of the relevant belief’s grounding. Its principal challenge is to accommodate sincere difficult intentions under adverse evidence without either redefining the content as trying or weakening “belief” until the identity thesis loses its bite (Marušić & Schwenkler, 2018).

Practical-knowledge theories explain immediate self-knowledge of action under way and the distinctive practical standpoint from which the agent understands what she is doing. Their challenge is to identify exactly what is known when the act is incomplete, interrupted, or misguided. Such theories gain precision by distinguishing the progressive description of an activity from its completed result (Anscombe, 1957/2000; Setiya, 2016).

Partial-belief approaches explain the variability of confidence and the persistence of commitments despite uncertainty. Their task is to show how degrees of confidence bear upon rational intention without reducing the commitment to a mere credence threshold. An action can be worth adopting despite a low prospect of success, and an event can be highly probable without being wanted or intended (Holton, 2008).

These positions can be tested against common constraints. A useful theory must preserve the agent’s practical authority, explain means–end organization, accommodate failure, distinguish prediction from resolve, respect epistemic evidence, and specify when self-knowledge is genuine. Different theories achieve these tasks through different conceptual mechanisms. Their comparative value lies in how well those mechanisms fit the full range of cases, including those where sincere agency and rational uncertainty coexist.

Interdisciplinary Implications and Attribution of Agency

The debate has implications for cognitive science, decision theory, and the interpretation of goal-directed systems. Psychological models can distinguish predictive confidence, goal commitment, motor planning, monitoring, and control. Such empirical distinctions are relevant because philosophers may otherwise impose a single attitude category upon separable mechanisms. Yet a difference in cognitive processing does not by itself prove the metaphysical identity or nonidentity of intention and belief.

In decision theory, an agent may choose an action with a low probability of achieving a valuable outcome because expected value, risk distribution, or moral importance favors trying. This shows that action selection need not be equivalent to prediction of successful achievement. It does not show that every philosophical intention can be represented by a utility-maximizing decision rule. The intentional object and the agent’s practical reasons still require interpretation.

Artificial and distributed systems raise a related issue of attribution. A planning system can represent goals, estimate success probabilities, revise forecasts, and select means. These operations make the philosophical distinction between commitment and prediction useful for describing its public organization. Whether such a system literally bears beliefs or intentions depends on a substantive theory of mentality, agency, and semantic attribution. Observable goal-directed output alone underdetermines that question.

The central philosophical result applies across the investigation without requiring a separate theory of artificial agency: explanations become more precise when the practical role of an adopted goal is distinguished from the epistemic status of a forecast. This distinction can guide research on human action, collective projects, software agents, and other organized systems, while questions about consciousness and metaphysical subjectivity remain separate matters.

Open Questions: Entailment, Warrant, and Control

Several problems remain open. The first is whether there are genuine intentions with an explicit contrary full belief and, if so, whether they are rational or merely possible. The answer depends on the individuation of the intended act, the time at which the attitude is assessed, and whether the believed impossibility concerns the act itself or an external outcome.

The second is how intention affects rational confidence. Does it generate positive evidence through the agent’s control, create a practical rather than theoretical entitlement, or impose a rational requirement to reorganize one’s beliefs? A comprehensive account needs to explain why these effects vary with competence and circumstances, rather than treating all deliberate commitments as equally reliable.

The third concerns knowledge of incomplete action. Can an agent know what she is doing when success depends on conditions not yet realized? The imperfective account gives a promising answer for ongoing activity, but it also needs principled limits: a failed attempt cannot retroactively make every intended result a completed intentional deed.

The fourth concerns the source of means–end norms. Cognitivist reductions and planning explanations illuminate different aspects of practical coherence. Further argument is needed to establish whether an epistemic norm can fully explain the demand to adopt means, or whether practical commitment introduces a distinctive normative relation.

The fifth concerns attribution beyond the familiar human case. If intending requires belief, any extension of intention to nonhuman agents inherits requirements concerning belief content and epistemic standing. If intending is primarily a planning commitment, other kinds of purposive organization may qualify under different conditions. The question is conceptual and evidential, demanding clear criteria before conclusions about particular systems are drawn.

Philosophical Synthesis: Practical Commitment and Epistemic Expectation

The historical reconstruction establishes that the distinction between practical orientation and cognitive representation has deep antecedents. Aristotle separates deliberative choice from opinion. Aquinas locates intention in the will’s directedness toward an end while preserving its rational organization. Anscombe articulates practical knowledge and the normative form of intentional descriptions. Davidson shows how belief enters causal rationalization without simply identifying every practical state with a prediction. Bratman explains the organizing functions of intentions in plans; Velleman and strong cognitivists investigate the special practical role of beliefs about one’s own conduct.

A defensible synthesis must preserve four relationships. First, intention has a constitutive practical role: it settles what the agent is to do and structures future action. Second, beliefs about circumstances and means are indispensable to the rational management of intentions. Third, an intention often provides a reason to expect a corresponding act, because a competent agent’s settled commitment is ordinarily part of the causal and normative explanation of performance. Fourth, this expectation can fall short of all-out belief, particularly where risk, disability, other agents, or uncertain circumstances limit control.

This yields a conditional epistemic principle: the stronger an agent’s effective control of an intended act under its relevant description, the better intention can function as a warrant for expecting that act; as control becomes indirect or incomplete, the rational relation between intention and unqualified success belief weakens. This is a philosophical synthesis advanced by the present monograph, not a newly declared canonical theory or a claim of historical priority.

The principle does not equate rational confidence with the probability of success alone. The practical object must be correctly individuated. Someone may know that she is attempting a rescue, remain committed to successful rescue, and rationally expect substantial risk of failure. Those three attitudes concern different but connected contents. Their coexistence is intelligible once commitment, ongoing action, and achievement are distinguished.

Belief and intention are therefore connected through an organized structure of reasons, capacities, expectations, means, and performance. A universal law that every intention entails categorical belief in successful completion fails to respect this structure. A theory denying every epistemic consequence of intention fails for the complementary reason: it overlooks how one’s deliberate conduct can legitimately inform what one takes oneself to be doing and likely to do. The middle ground is principled because it specifies the conditions under which commitment supports knowledge or justified expectation.

Conclusion

The relation between intention and belief is a problem about the unity of practical agency and epistemic accountability. An agent can determine what to do, know what she is doing, and forecast the result, but those activities need not have identical objects or standards of rationality. The history of the debate reveals the persistent attraction of treating intention as belief about one’s future conduct and the equally persistent resistance generated by uncertainty, partial control, and practical commitment.

The central analytical result is that intentions ordinarily govern performance and organize beliefs about means, while beliefs about success depend upon the agent’s competence, available evidence, degree of control, and the intended action description. Practical knowledge of action in progress explains some first-person authority without extending that authority automatically to every future outcome. Planning theory explains the stability of intentions without divorcing commitments from epistemic expectations. Strong cognitivism identifies an important class of practically grounded beliefs, while its claim of general identity remains contested.

Intention to act, intention to try, belief that one will act, and knowledge of acting occupy distinguishable positions in one purposive process. Understanding their relations resolves several apparent contradictions and provides a more exact account of rational action under uncertainty. The enduring philosophical question is how practical determination can generate epistemic authority while remaining answerable to the genuine possibility of failure.