Action Guidance and Control

The Question of Guidance in Action

The philosophical problem of action guidance concerns what makes an unfolding performance an exercise of an agent's control. An intention can settle what someone will attempt, and a movement can bring about the intended result, while the connection between them remains philosophically obscure. Consider a person turning a key to open a door. The initial decision identifies a goal, but performing the action requires selecting the key, orienting it, adjusting the wrist when resistance is encountered, recognizing the movement as relevant to opening, and abandoning or revising the attempt if the lock is broken. Guidance is the temporally extended organization by which conduct continues to answer to a practical end. Control is the agent's capacity, manifested in appropriate conditions, to keep the course of conduct responsive to that end.

The problem therefore arises within an action, not merely before it. A prior intention does not itself specify every muscle contraction or alteration of force. The practical significance of a movement cannot always be read from its physical description. An identical displacement of the hand might be a correction, an accidental twitch, a response to threat, or an irrelevant movement. To distinguish these possibilities, an analysis must attend to the agent's settled or emerging goals, the action's environmental conditions, and the relationships among initiation, monitoring, correction, and completion. Aristotle's discussion of the origin of voluntary movements, Anscombe's account of knowledge in intentional action, Frankfurt's critique of purely antecedent causal analysis, Bratman's planning theory, and recent work on motor control approach different portions of this problem (Aristotle, Nicomachean Ethics, III.1–5; Anscombe, 1957/1963; Frankfurt, 1978; Bratman, 1987; Shepherd, 2021).

The central thesis developed in this monograph is that guidance is an action-relative relation among practical orientation, ongoing sensitivity to relevant information, and a structured capacity to vary or sustain performance. Control is assessed by the appropriate counterfactual responsiveness of that relation: what the performer can do when the circumstances change in ways that matter to the task. This proposal is analytical rather than a new canonical theory. It is meant to specify the explanatory work demanded of competing philosophical accounts, while leaving room for the causal, teleological, phenomenological, and embodied explanations of how the relation is realized.

A successful action is evidence relevant to control, but success alone is inconclusive. A novice who hits a target by accident and an expert who intentionally chooses an extremely difficult shot can have different degrees of control despite identical outcomes. A surgeon may exercise substantial control even when an intervention fails because of an unexpected condition; a driver may reach the destination while repeatedly losing the capacity to respond to traffic. Control concerns how the performance is governed across possibilities, not just whether its final state coincides with a prior desire. This distinction supplies the monograph's guiding test for accounts of intending, acting, and accomplishing.

Definition and Philosophical Scope

Action guidance is the organization of an ongoing performance by a practical end under conditions in which information about the performance or its circumstances can shape what happens next. Such information may be consciously attended to, habitually exploited, perceptually structured, or incorporated by trained motor processes. Control is a capacity to maintain, revise, redirect, inhibit, or terminate conduct in ways relevant to the agent's practical orientation. The agent can possess control without exercising every available adjustment, and a single exercise can exhibit control without success. Shepherd's distinction between possession of control and its exercise provides an important contemporary precedent for treating capacity, manifestation, and outcome separately (Shepherd, 2021, chs. 2–4).

A practical end has a determinate function in a particular episode. For a pianist it may be sustaining a passage, for an editor detecting an error, and for a person escaping a fire reaching safety. Ends can be nested: pressing a pedal serves a musical phrase, and the phrase serves a performance. They can also change as circumstances change. Guidance thus has an evaluative aspect even when an action is morally objectionable. A deceptive speaker may control the timing of an utterance with great skill. Moral justification and effective control are different assessments, although the ability to control action often bears on responsibility.

Several distinctions set the research boundary. Intention is a practical commitment or orientation; execution is the realization of the attempted course; guidance describes what coordinates its development; control concerns the capacity exercised in making that coordination effective; success describes a satisfactory outcome relative to an end; and the sense of agency concerns how agentive performance is experienced or attributed. The boundaries are porous in actual action, yet their conceptual separation prevents a theory of one phenomenon from silently substituting for a theory of another. A performance can be intentionally initiated and poorly guided. It can be guided by a skillful process that the agent cannot describe in explicit propositions. It can be experienced as self-produced even when its causal basis is disputed.

Guidance does not always require an explicit command from reflective deliberation. Skilled handling of a bicycle often consists of rapid adjustments that could be disrupted by painstaking verbal monitoring. Conversely, some activities, including hazardous technical work, call for explicit checking and reconsideration. The research question is how both sorts of performance can remain genuinely agentive. An adequate theory must avoid equating personal-level agency with uninterrupted conscious micromanagement while explaining how subpersonal processes can serve reasons and ends attributable to an agent (Fridland, 2017, 2021).

The present dossier concentrates on the connection between intention and execution, feedback, error sensitivity, adaptive correction, performance stability, and the criteria of control. Neighboring research subjects include intention itself, reasons for action, practical commitment, skilled action and habit, free will, and agency's phenomenology. These neighboring questions enter when they illuminate guidance. Their independent histories and full theoretical landscapes belong to their own research dossiers.

An Explanatory Architecture of Guided Performance

The first task is to identify the explanatory levels at which one can ask whether conduct is guided. At the level of practical specification, an agent organizes a task through purposes, conditional commitments, and priorities. At the level of performance organization, the agent selects movements, sequences subgoals, and preserves attention to task-relevant features. At the level of execution, bodily or technological processes implement selected adjustments. At the level of appraisal, information about results and unexpected conditions changes subsequent conduct. These are analytic roles, not an assertion that four independently localized psychological modules exist.

A requirement for guidance at one level may be satisfied by a different mechanism at another. A person can deliberately select a route and then navigate individual steps without renewed choice; a trained climber can adjust grip pressure before articulating a reason; an operator can direct a remote device whose local controller regulates forces unavailable to human perception. Agency can therefore extend across nested processes. The philosophical question becomes how the lower-level regulations remain integrated with the wider course of action rather than becoming causally insulated behavior merely associated with a prior plan.

Three characteristic structures are especially important. Prospective guidance constrains the range of actions considered or prepared: I decide to travel by train and therefore look for departure times. Online guidance regulates execution as it unfolds: I slow my pace when the platform is crowded. Retrospective guidance affects subsequent attempts: I recognize a navigation mistake and alter tomorrow's route. These temporal modes belong to a unified practical activity when the information used at each point bears on what the agent is trying to accomplish. Neither feedback nor learning needs to take the form of a spoken judgment.

A useful comparison is between a thermostat and a person adjusting a room for guests. Both can exploit discrepancies between a measured temperature and a reference value. The person's action may involve deciding whether warmth, energy consumption, ventilation, or a guest's discomfort matters more, and may change the target itself. This illustrates why feedback regulation is an important possible component of action guidance but does not exhaust the philosophical idea of agency. Engineering descriptions of regulation illuminate mechanisms. A philosophical account additionally interprets practical authority, reasons, standards of relevance, and the individuation of the action under those standards.

The framework also makes room for open-loop execution. A practiced movement may proceed rapidly with little opportunity to use external sensory feedback during its shortest phases. Such movement remains guided when prepared motor organization, prediction, and capacity for modulation link the performance to the task. The philosophical classification should be sensitive to the temporal scale of the action, rather than stipulating that every instant of agency must contain a newly completed feedback loop (Wolpert, Ghahramani, & Jordan, 1995; Miall & Wolpert, 1996).

Aristotle: Origin, Voluntariness, Choice, and Practical Movement

Aristotle's analysis of voluntary action offers an early account of the agent's relation to the source of performance. In Nicomachean Ethics III.1 he distinguishes acts done under compulsion or through ignorance from voluntary acts and treats the location of the moving origin as decisive when the agent contributes nothing to the event. The examples include being carried by wind or another force. Yet Aristotle also considers actions under severe pressure that are chosen in the circumstances, such as jettisoning cargo during a storm. An agent can retain a practical role in guiding performance under constraints that make the overall situation undesirable (Nicomachean Ethics, III.1, 1110a1–1111a24; English text in Ross's translation and Greek text in Bywater's critical edition).

Aristotle's Greek terms matter. Archē (ἀρχή) concerns an origin or principle; prohairesis (προαίρεσις), often translated as choice or decision, names a distinctive form of deliberate practical commitment. Boulesis (βούλησις) identifies wish, while bouleusis (βούλευσις) denotes deliberation. In Book III Aristotle distinguishes deliberation about what is in our power from wishing for ends, and characterizes choice as involving deliberative desire about what is practicable. This complex family of concepts cannot be collapsed into a modern decision-versus-act switch. It already connects the circumstances of performance, the agent's practical origin, and the selection of means (III.2–5, especially 1111b–1113b).

The De Motu Animalium and De Anima place the relation between thought, desire, perception, and animal movement at the center of natural philosophy. Aristotle recognizes that perceived and imagined objects can move animals through desire; a practical conclusion may be realized in movement rather than remain a detached speculative judgment (De Anima III.9–11; De Motu Animalium 6–7). Translating his account into a contemporary architecture of motor control would be anachronistic. Its philosophical contribution lies in asking how cognition and appetite, oriented to something as good or desirable, have an intelligible place in the production of movement.

His treatment also introduces a methodological complication. The same visible behavior may be deliberate or forced depending on its genesis, while the same agentive origin can support a range of imperfect outcomes. Yet Aristotle's voluntary/involuntary distinction answers some questions of imputability that differ from whether execution was skillful or sufficiently responsive. The contemporary project must preserve that difference. A person can voluntarily initiate an action while being unable to guide it precisely. Voluntariness explains an important condition of ownership and responsibility; control analysis further asks which variations of performance the agent could appropriately make.

The Aristotelian inheritance is therefore twofold. The agent's contribution is constitutive of the philosophical characterization of action, and acting is intelligible through the end for the sake of which means are selected. Both themes survive in contemporary causal and teleological theories even when their metaphysical accounts of movement differ.

Will, Motor Learning, and the Nineteenth-Century Problem

William James's Principles of Psychology (1890), Chapter XXVI, develops a psychological picture in which an idea of a possible movement acquires motivational significance from the agent's experience of movement. James argues that a subject must have some experiential basis for anticipating a movement before voluntarily reproducing it. His account of ideomotor action connects the anticipated effects of movement to the execution of bodily responses. It is historically important because it identifies a gap between representing an end and possessing the learned capacity to produce the appropriate means.

James gives attention an important role in difficult willing: maintaining a practically relevant idea against competing tendencies can change the action that follows. His account addresses inhibition, effort, the acquisition of voluntary movement, and failures in translating a stable practical aim into bodily activity. The full text shows both James's ambition to explain voluntary movement psychologically and his explicit uncertainty about the neural mechanisms by which a prevailing idea activates motor processes (James, 1890, vol. 2, ch. XXVI, especially pp. 486–488 and 559–565). The historical lesson is that a philosophical account must distinguish the agent's determination from the enabling mechanisms of its execution.

John Dewey's “The Reflex Arc Concept in Psychology” (1896) criticizes the treatment of stimulus and response as detachable elements whose simple linear sequence explains conduct. His alternative emphasizes coordinated activity in which sensory and motor phases receive their functional significance from the ongoing adjustment. Dewey does not thereby offer a contemporary mathematical feedback-control theory; he contests an explanatory abstraction in which an organism's action is decomposed without its purposive organization. The relation between perception and movement is already dynamic, and what counts as a relevant stimulus depends on the activity being carried out (Dewey, 1896, pp. 357–370).

Taken together, James and Dewey press complementary questions. How is a conceived act made executable through acquired capacities? And how do perceived circumstances and motor activity acquire their functions within coordinated performance? The first leads toward learning, intention, and action selection; the second toward embodiment, context, and the organization of sensorimotor loops. Contemporary philosophy inherits both pressures when it connects personal-level descriptions of agency with experimentally informed motor models.

Ryle, Phenomenology, and the Intelligence of Execution

Gilbert Ryle's The Concept of Mind (1949), particularly its treatment of knowing how, contests the picture in which intelligent performance must be accompanied by an indefinitely repeated sequence of internal theoretical instructions. Knowing how manifests itself in discriminating circumstances, modifying conduct, and exercising appropriate abilities. This does not establish that skilled action lacks thought, nor does it settle the contemporary debate over propositional accounts of know-how. It does show why the presence of silent commentary cannot be the universal condition of guidance. If each intelligent adjustment required a further deliberate instruction, the account would postpone the source of intelligence indefinitely.

Maurice Merleau-Ponty's Phenomenology of Perception (1945; Landes translation, 2012) develops the concept of motor intentionality in the setting of embodiment and bodily orientation. Ordinary action has a practical directedness that exceeds explicit judgment. The skilled body is attuned to an environment of possible movements; reaching, grasping, and navigating can reflect a lived understanding of how to act. His discussion of Schneider, pathological movement, and bodily habit is especially relevant to the dissociation between capacities of abstract movement and contextually organized performance. The concept of motor intentionality identifies a distinctive experiential and practical organization; it should not be equated without argument to present-day technical definitions of motor representations.

These traditions make a shared challenge to accounts that reserve genuine guidance for operations describable as explicit rational calculations. An expert can know what she is doing in ways revealed by timing, discriminations, and responsive adaptation. Yet one must not infer that every automatic behavior counts as the exercise of agency. A startle reflex, a trained habit maintained against a decisive change of goal, and a fluid correction of a skilled movement have different practical relations to the agent's situation. The relevant inquiry concerns integration, flexibility, and appropriateness, not a categorical opposition between conscious reasoning and motor automaticity.

The strongest phenomenological insight is that the organization of the bodily situation can already have significance for action. The strongest demand from analytic philosophy is to explain when and how such significance constitutes acting by an agent rather than purposive behavior merely occurring in an organism. A mature account of guidance benefits from both questions while keeping their respective criteria visible.

Anscombe: Knowledge in Action and the Practical Question

G. E. M. Anscombe's Intention (first published 1957; second edition 1963) changes the philosophical point of departure. Intentional action is investigated through the distinctive applicability of the question “Why?” and the descriptions under which an agent can answer it by giving a reason. A person may knowingly move an arm in order to pump water, which in turn serves a further end. The description under which an action is intentional is central: someone can intentionally move the arm without knowing that doing so has an unnoticed secondary effect (Anscombe, 1963, §§5–26).

Anscombe's treatment of practical knowledge challenges the assumption that an agent knows what she is doing solely by observing herself as an external spectator. In normal intentional action, knowledge of what one is doing is bound to the practical organization of the deed. Her famous discussions of the shopping list and the direction of fit between practical and speculative knowledge locate a kind of error in performance rather than judgment when conduct fails to conform to the intention (Anscombe, 1963, §§32–33, 45–48). The agent's specification of the action provides a standard by which the unfolding performance can be assessed.

This does not imply that agents are infallible. One may sincerely say that one is writing a letter while an unexpectedly failed input device records nothing; one can be mistaken about whether a sequence is being executed. The philosophical challenge concerns the extent to which the special authority of practical knowledge remains even when success or perceptual knowledge is limited. Contemporary discussions differ over whether such knowledge is constitutive of intentional action, whether it is supported by perceptual information, and whether it needs an account of reliable self-knowledge (Setiya, 2007; the Stanford Encyclopedia of Philosophy entry “Intention,” revised 2022).

For guidance, Anscombe's crucial contribution is to make the content of the action intelligible from within the practical reasoning that organizes it. A mere match between an outcome and an external prediction leaves out what the person is doing under a reason-giving description. Yet an Anscombean account must still explain the detailed bodily and temporal realization of the activity. Knowing that one is preparing tea need not specify how the hand compensates for a cup's shifting weight. The epistemological account of action and the organization of its execution are mutually constraining research programs.

From Intentional Descriptions to a Theory of Control

The historical traditions yield a precise research problem. Aristotle asks where and how the agent originates action; James examines the acquisition of voluntary movement and effort; Dewey situates responses in coordinated activity; Ryle identifies intelligent competence in its exercise; Merleau-Ponty analyzes embodied practical orientation; and Anscombe connects action to its reason-giving description and practical knowledge. None of these questions can be answered simply by establishing that a bodily movement occurred after a desire or intention.

An integrated theory must explain why an agent's relation to a performance persists through its duration. It must account for the possibility that action guidance is distributed across several timescales, from advance planning to rapid motor adjustment, and for cases where guidance is lost while motor movement continues. It must also explain how error can be detected and responded to without making external success the sole test of control. The central modern controversy turns on whether a causal sequence from appropriate psychological antecedents suffices, whether a special process of guidance is necessary, and how intentions, skills, and environmental information sustain that process.

The following chapters reconstruct those controversies, beginning with Davidson's causal account and the problem of deviant causal chains, and moving to Frankfurt, planning theory, practical knowledge, control capacities, and contemporary motor science. This order is argumentative: each position supplies part of the explanation that the prior one leaves open, but the inquiry preserves their genuine philosophical disagreements.

Davidson: Reasons as Causes and the Individuation of Action

Donald Davidson's “Actions, Reasons, and Causes” (1963) proposes that the primary reason for an intentional action explains it by rationalizing the action and standing in the relevant causal relation to it. A primary reason combines an attitude favorable to a certain kind of conduct with a belief about how the action contributes to it. If someone turns a switch to illuminate a room, a wish to make the room light and a belief about the switch may explain why the person acts. The explanatory role of reasons includes a causal claim; otherwise the reasons cited might be reasons the agent had but which did not produce the action in question (Davidson, 1963, pp. 685–700).

This is a major advance in the analysis of guidance because it distinguishes a rationalizing description from mere temporal coincidence. The reason must matter to the occurrence of the action. Davidson's event ontology further allows one event to admit several descriptions. Pressing a switch, turning on a light, and alerting someone by illuminating a room may identify the same action under differing descriptions, although the agent need not intend every consequence. The philosophical difficulty is to identify those descriptions whose explanatory relation to the agent's reason constitutes acting intentionally. The existence of a physical event alone does not answer the question.

Causalism is not committed to the crude claim that every desire followed by a matching bodily movement constitutes intentional action. Davidson himself acknowledges the problem of deviant causal chains: an agent can have the appropriate reason and that reason can cause an outcome through an unexpected route that does not amount to the relevant intentional deed (Davidson, “Freedom to Act,” 1973; Essays on Actions and Events, 1980). A reason may produce nervousness that accidentally loosens a grip; a plan to signal a friend may cause such excitement that the hand inadvertently makes the intended gesture. In each case the outcome can match what was wanted and can be causally dependent on wanting it, without being controlled as intended.

The formal issue is not whether a sequence contains a cause. It is whether the causal history is appropriately action-constituting. Let R designate a reason, B a bodily movement, and G the guidance relation required for the practical performance. The naive sufficiency claim, R causes B, therefore B is intentional, fails. A more sophisticated thesis requires that R cause B through a course in which G is realized. Yet if G is defined simply as the course appropriate to an intentional action, the theory risks circularity. It must say what makes the course appropriate: sensitivity to relevant changes, exercise of competence, integration with a sustained plan, or a different noncircular criterion.

Davidson's theory retains considerable force. Reasons can provide genuine causal explanations, and the movement's relation to reasons helps individuate its intentional description. The challenge is that causation considered without the action's organization underdetermines guidance. His account supplies a necessary explanatory dimension for many actions while leaving open the philosophical characterization of the path.

Frankfurt: Guidance Within the Performance

Harry Frankfurt's “The Problem of Action” (1978) directly contests the strategy of locating the entire distinction between what agents do and what happens to them in antecedent causal history. His argument turns attention to the nature of the ongoing performance. If agents are active in what happens, their activity must be distinguished from mere bodily happenings by a feature of the process itself. Guidance, correction, and the relation of a movement to the agent's direction become central explanatory candidates (Frankfurt, 1978, pp. 157–162).

Frankfurt's intervention should be read carefully. He does not show that causes are irrelevant to action, and a defender of causalism can respond by giving a more detailed account of causal organization. The point is that the right cause, identified solely as an antecedent belief, desire, or decision, may not explain why an event is under the agent's control throughout its unfolding. A person whose arm is mechanically carried through a motion may have wanted that motion and may have initiated circumstances that cause it, yet not be guiding its subsequent trajectory in the way characteristic of free movement.

Guidance is therefore a process requirement. An action must retain a form of subjection to the agent's direction. Frankfurt's proposal raises further problems: what exactly constitutes this subjection, how can it be identified in cases of highly automatized skill, and can a system guide activity without reflective awareness? These questions later accounts of control attempt to answer. If guidance is equated with whatever happens in genuine action, it supplies a label rather than a criterion. If it requires continuous conscious monitoring, it risks excluding much competent performance.

The epistemic asymmetry is revealing. A spectator may know the outcome and the prior decision while lacking information about the manner of performance. The agent's competence and responsiveness can be revealed under perturbation. For example, two players may execute the same motion when the field is clear, but only one can adjust reliably when an opponent unexpectedly moves. The difference is counterfactual and processual. It supports the claim that control must involve a suitable capacity rather than the isolated historical fact of an intention causing a matching result.

The Frankfurtian challenge is best taken as an explanatory demand on theories of action: identify the ongoing structures by which an agent's practical orientation is expressed as performance. Causal theories that meet the demand remain live competitors. The ensuing debate concerns the nature of those structures.

The Problem of Causal Deviance

Deviant causal chains expose the gap between motivation and guidance. Consider a mountaineer holding another climber's rope. The mountaineer wishes to release the rope but is so disturbed by the wish that the hand involuntarily relaxes. The wish causes a release; the release achieves the desired end; still, this is not the intentional releasing originally entertained. Davidson's example has become a standard touchstone for action theory (Davidson, 1973; Action, Stanford Encyclopedia of Philosophy, 2023, §2.4).

A second kind of deviation can arise after apparently intentional initiation. A person intentionally operates a mechanism to produce a signal, but an unrelated accident makes the signal occur through a route the person cannot regulate. The initial movement may be intentional while the distal outcome is not an intended achievement in the relevant manner. These cases show that distal success and local guidance must be individuated with care. At which point does the intentional action terminate, and which later consequences are attributable to it? An answer requires the practical description under which the event is assessed.

A general counterfactual test begins with a question: if the task-relevant conditions had changed within an appropriate range, would the performance have altered in a manner responsive to the agent's end? If the mountaineer were alerted to the rope's weight, would the hand's grip have been adjusted through an exercise of the intended act, or would the nervous spasm have proceeded independently? This test is useful but must be qualified. A highly constrained environment can hide control; a fortunate chance sequence can mimic responsiveness in a single episode; and counterfactual performance depends on the difficulty and temporal scale of the task. Reliable patterns across relevant possibilities matter more than one imagined alternative.

Nondeviance can accordingly be approached through the organization of the action's causal process. A goal-directed system can constrain options, recruit suitable means, monitor discrepancies, and correct or terminate execution. The agentive status of that organization depends on its relation to the person or other candidate agent's practical orientation, rather than on the mere existence of a technical feedback loop. An independently operating device could compensate for disturbances even if the person who activated it did not exercise detailed control over its subsequent movements.

A philosophical theory must also distinguish deviance from normal uncertainty. Successful intentional action often depends on conditions outside the agent's control. A driver cannot determine whether a road is unexpectedly blocked, but may still navigate competently if the obstruction appears. The fact that success depends on luck does not automatically imply that the driver's performance was uncontrolled. The crucial issue is what components of the process are appropriately responsive to what the agent can perceive, predict, or practically manage.

Bratman: Plans as Structures of Guidance

Michael Bratman's planning theory explains intentions partly through the characteristic functions they perform in temporally extended agency. Intentions are elements of partial plans, coordinating future behavior, constraining reconsideration, guiding means-end reasoning, and organizing action across time and among persons (Bratman, Intention, Plans, and Practical Reason, 1987). A plan enables agents with limited deliberative resources to secure coherent courses of conduct without deciding every detail at every moment.

The transition from future-directed intention to action is a central contribution of this view. If I intend to travel tomorrow, my intention makes arranging transportation relevant and rules out certain incompatible alternatives unless I reconsider the plan. Yet the plan is usually incomplete. I need not decide now which exact train car to enter or which path through a crowded station I will take. Partiality is a resource: it leaves spaces for information-sensitive completion. Guidance requires both stability of commitment and responsiveness to circumstances.

Bratman's treatment of means-end coherence helps explain why an intention has rational implications for execution. When an agent intends an end and understands that adopting a means is required, unresolved gaps in the plan generate practical pressures. But one should distinguish rational pressure from causal guarantee. An agent might fail to deliberate well, might form inconsistent intentions, or might lack the physical ability to execute a coherent plan. The planning theory describes normative and organizational functions of intentions; a complete theory of performance also needs an account of how capacities implement them.

Bratman's discussion of reconsideration is especially relevant. Constantly reexamining an intention would destroy the benefits of planning, while rigid adherence in radically altered circumstances would undermine agency. Rational stability is defeasible. Guidance is therefore not equivalent to stubborn persistence. The agent who changes course when a bridge collapses may remain more faithful to the point of the plan than one who continues blindly toward the blocked crossing. What is retained is an organized relation to the practical end, not an unvarying sequence of physical operations.

The theory yields a structural insight: intention guides by shaping what counts as a relevant next step. An agent following a plan does not need an explicit representation of every motor act, but the plan must constrain the local forms of activity that count as implementing it. In complex performances, that constraint is distributed among standing intentions, conditional subplans, habits, tools, and perceptual opportunities. Bratman's framework provides a theory of how the personal-level commitment organizes a task. Questions about motor representation, skill, and embodied control investigate how the organization becomes executable.

Proximal Intention, Initiation, and Sustaining Conduct

Intentions can function at different temporal distances from behavior. A distal intention concerns a future project; a proximal intention concerns an imminent action; an intention in action is sometimes understood as the active direction of performance. This vocabulary is widely used in philosophical psychology, but it should not be treated as three independently proven faculties. It distinguishes explanatory roles in complex conduct (Mele, 1992; Pacherie, 2008).

Alfred Mele emphasizes executive dimensions of intentions in Springs of Action (1992), analyzing how intentions relate to beliefs, desires, decisions, and the initiation of behavior. A person may intend to rise at six, then decide to begin rising, and then actually move. The sequence highlights a genuine gap: commitment can be present while action initiation is blocked. In other cases, an action begins and must be sustained against distraction, fatigue, or environmental resistance. The ability to initiate is thus one dimension of control; continuation and calibration are others.

The distinction becomes salient in akrasia, procrastination, and weakness of will. An agent can judge that exercising is preferable and sincerely resolve to exercise, but spend the evening doing something else. This case is often a failure of commitment formation, maintenance, or action initiation rather than a technical inability to execute the exercise movements. Another agent can begin an exercise deliberately and lose control of pace in ways that defeat the intended regimen. Conflating these failures erases the mechanisms and norms of guidance relevant to each.

A philosophical analysis benefits from distinguishing intention's motivational efficacy from its informational specificity. A strong commitment may be poorly specified: “repair the device” offers little immediate guidance when the agent lacks procedural knowledge. A comparatively unreflective intention can support precise execution when competence is well established. Guidance is an achievement of the coordination between practical orientation and available ways of acting; intensity of desire or certainty of decision cannot substitute for the competence that makes the task executable.

This point has an implication for the concept of will. At times “control yourself” means suppressing a competing urge or maintaining a prior decision. At other times it means attending to a delicate sequence of operations and responding to changes. These are both agentive achievements, but a theory explaining one does not automatically explain the other. The present dossier focuses on their common structural feature: the agent's conduct remains answerable to a task-relevant practical orientation over time.

Guidance, Reasons, and the Standards of Correctness

A reason for action can justify a course of conduct, explain its occurrence, or help organize the selection of means. These functions must be distinguished in analyzing control. A person may have good reason to perform an action but execute it incompetently, or perform a morally poor action with exquisite technical control. The normative question of whether the end is worth pursuing differs from the question of whether conduct is effectively guided toward it. Nonetheless, reasons help identify which outcomes and corrections matter for a performance.

A person walking home because of threatening weather treats shelter as an end and changes route when a street floods. The change is intelligible as a correction because it serves the original purpose. A movement that follows the same route change accidentally could achieve the same result without being guided. What makes the correction a practical correction is its dependence on relevant information and on the continuing role of the end. Explanations at the level of reasons specify that dependence; motor and causal explanations specify how the dependence becomes effective.

Anscombe's practical-knowledge approach supplies an account of standards internal to intentional action. If one intends to place a glass on a table and instead drops it, the relevant mistake may be a failure of execution rather than a false judgment about what would be desirable. Yet some errors can undermine self-knowledge: an agent might be unaware that the glass has already shattered. Theories differ about whether intentional action entails current knowledge of what one is doing, an ability to know, or justified reliance on a practical representation. Those alternatives have different implications for guidance under perceptual uncertainty (Anscombe, 1963, §§45–48; Setiya, 2007).

The standards of correction may change during action. If the table collapses, the agent can redirect the glass to another surface without abandoning the end of placing it safely. A rigid description such as “put the glass at point X” may no longer serve the purpose under which the original plan was adopted. A theory limited to compliance with initial commands overlooks practical intelligence in revising the specification of success. The test of control should therefore include the agent's authority to reconsider means and, in some circumstances, the end itself.

The relation of reasons to motor detail is asymmetric. The agent may know why a glass should be set down gently without knowing the exact motor forces used. Conversely, a motor prediction can specify the force needed to accomplish a movement without deciding why the task is worth doing. Neither explanation replaces the other. Guidance joins them in a relation of task-sensitive implementation.

A Comparative Appraisal of Causal and Teleological Accounts

Causal theories emphasize the explanatory connection between appropriate psychological conditions and action. Teleological theories emphasize intelligibility in relation to an end. Frankfurtian approaches stress guidance in the unfolding process. Planning theories explain stability and revision of practical commitments. Each view can explain a different feature of successful agency; their disagreements concern which features are essential and how they are related.

Rowland Stout's Things That Happen Because They Should (1996) offers an explicit teleological approach to action. His claim concerns explanation in terms of what an activity is for, including rational or practical evaluation, and is not a claim that all actions are morally good. Teleology can identify what makes a sequence into an intelligible activity, while a causal explanation can specify how the sequence was produced. The unresolved philosophical issue is whether one can derive the action's teleological unity from purely mechanistic facts without presupposing the intentional description that is to be explained.

The causal objection to teleology is that invoking ends may merely redescribe a pattern already produced by ordinary mechanisms. The teleological objection to reductive causalism is that causation by antecedent states may leave the intentional organization of the activity unexplained. A constructive response distinguishes constitutive and productive questions. What makes the sequence count as the execution of this task? Through which causes did the sequence occur? A complete theory seeks answers to both, and it should not assume in advance that their explanatory vocabularies are interchangeable.

It is a further question whether guidance requires a single unified controller. Complex agents can recruit routines, external instruments, and social assistance. If the agent remains capable of correcting or withdrawing authorization, delegated processes may be components of intentional action. At the same time, an uncontrolled automated consequence need not inherit the intentional status of every event that caused it. A theoretical account must identify where practical direction persists and where it has been displaced.

The philosophical synthesis advanced here treats causal production, practical organization, and counterfactual control as related but distinguishable explanatory dimensions. A pattern of behavior can be causally attributable to a reason without being controlled in the relevant way. It can be instrumentally successful without conveying the agent's continued practical authority. And it can express genuine control despite falling short of its intended goal under adverse conditions. These distinctions establish the agenda for contemporary theories of embodied and technologically mediated performance.

Control as a Capacity and Its Exercise

The contemporary debate gains clarity by distinguishing possession of control from its exercise. An agent may possess the ability to perform an action reliably under a range of suitable circumstances yet fail on a particular occasion. Another person may get the desired result by luck without possessing the relevant ability. Joshua Shepherd's The Shape of Agency (2021) develops this issue by treating control's possession in relation to reliability, psychologically guided behavior, and circumstances, and by connecting the analysis to nondeviant causation and intentional action. It supplies an important alternative to accounts that identify control with actual success.

The possession of control is relative to a task and a range of conditions. Someone may competently write with a keyboard but be unable to do so while wearing heavy gloves. A trained sailor may handle ordinary turbulence but not an unprecedented failure of the vessel. Specifying the circumstances matters because the unrestricted claim “the agent could always adjust” would exclude finite agents from genuine control. Competence is reliable across relevant conditions, not across every physically conceivable disturbance. The appropriate domain is partly determined by the ordinary practice, stakes, available information, and skill level involved.

Exercise adds an occurrence requirement. The agent's capacity must actually organize or regulate the performance, rather than be merely present as an unused possibility. A skilled tennis player who lets an automated machine swing the racket may possess tennis expertise without exercising it in the resulting stroke. A skilled player who deliberately performs a standard shot without conscious intervention can nevertheless exercise control through trained dispositions. The exercise of a capacity is not identical to conscious awareness of its operation.

Control is graded and multidimensional. An agent can control initiation but not timing, direction but not final impact, or general strategy but not the smallest variations in trajectory. A complete assessment should specify the action description, the feature of performance in question, the circumstances across which responsiveness is required, and the relevant level of control. The phrase “she controlled the action” is accurate only within such a practical context.

This analysis resists two temptations. One is outcome fetishism: treating every successful performance as evidence of excellent control. The other is mechanism fetishism: treating every feedback-regulating process as an exercise of agency. Capacity and exercise belong to a relation involving the agent, the action, and the task. Such a relation can be supported by multiple biological or artificial mechanisms.

Feedback, Prediction, and the Organization of Movement

Feedback control uses information about the state of a system to alter its ongoing operation in relation to a target. In bodily movement, sensory information can help an agent detect deviations and adjust. But the relevant mechanisms also include anticipatory or feedforward control. A rapid movement can be prepared through learned models of how action will affect bodily and environmental states. Work by Wolpert, Ghahramani, and Jordan (1995) provides experimental evidence relevant to internal models in sensorimotor integration, while Miall and Wolpert (1996) describe forward models that predict states from commands and current conditions.

The technical vocabulary requires precision. A forward model predicts the sensory or state consequences of a motor command; an inverse model, in some computational formulations, estimates commands appropriate to a desired change. Feedback refers to available information about actual state, whether proprioceptive, visual, auditory, or otherwise. Prediction can compensate for delays and support rapid control. These mechanisms are important empirical candidates for explaining bodily execution, but philosophical conclusions about agency do not follow merely from identifying them. A neural process can implement prediction without by itself constituting a reason-giving intention.

An elementary model clarifies the issue. Suppose a controller has a target state g, an estimated present state x, a policy for choosing an adjustment u, and feedback that updates its estimate. A discrepancy between g and x can inform a change to u. This description does not specify whose target g is, why it is relevant, whether it may be revised, or whether applying u counts as part of a deliberate performance. The model characterizes regulation. To understand agency, one must explain the relation between the regulated process and the practical activity attributed to an agent.

The temporal scale of feedback also matters. An archer cannot make a full visual correction to the arrow after release, yet may have exercised control over posture, timing, force, and direction before the release. An expert may perform rapid movements whose adjustment depends on prediction and prior training. The philosophical requirement is sensitivity across the relevant temporal organization of the whole action. A theory demanding fresh sensory correction at every moment would misclassify many normal and skilled actions.

The existence of prediction error should not be mistaken for the concept of an intentional mistake. In a computational model, the discrepancy may be mathematically defined relative to a reference variable. In intentional action, error also takes shape under a practical description: the note was too late, the signal was sent to the wrong person, the warning should have been withheld. These standards can be established or revised through reasons, institutional rules, or shared practices. Philosophical explanation must connect the levels without reducing practical correctness to the numerics of a control variable.

Motor Representations and the Interface with Intentions

One of the most influential interdisciplinary questions asks how personal-level intentions connect with motor representations that organize execution. Stephen Butterfill and Corrado Sinigaglia (2014) argue that purposive action presents an explanatory problem when intentions and motor representations have distinguishable roles. Standard action theory often emphasizes intentions as explanations of purposiveness, while motor science reveals representational structures with contents and functions that need not coincide with personal-level plans. Their challenge is to account for coordination between these forms of representation.

A representation of an intention might specify “open the drawer”; a motor representation may organize grip shape, direction, and force in relation to the handle. The contents differ in abstraction, timing, and format. Still, they must cooperate if the practical aim is to be realized. One possible explanation is hierarchical specification: a broader plan constrains which motor outcomes are useful, while local motor representations provide executable detail. Another is an embodied continuity account in which practical planning itself employs motor-modal structures, especially in skillful activity (Fridland, 2021).

The interface problem becomes acute in familiar cases. A person intends to take a fragile object without dropping it. A successful grasp recruits perceptual sensitivity to shape and weight, grip-force control, and a task-level commitment to preserving the object. If the motor system executes an efficient grasp that damages the object, successful motor optimization relative to a narrow parameter is not successful guidance relative to the person's wider practical end. The agent's relevant task description matters in evaluating the performance.

It is possible to describe the relation in both directions. Top-down constraints from practical intention can organize which opportunities and movements are selected. Bottom-up information can prompt revision of the means or reconsideration of the end. The phrase “top-down” is schematic: it does not entail a simplistic neural hierarchy or conscious commands to each muscle. Equally, “bottom-up” information may already be structured by learning, expectation, and practical relevance.

A complete philosophical interface theory must explain at least four things. It must characterize the contents and functions of the representations; identify the mechanisms by which they cooperate; explain how their cooperation constitutes the agent's performance; and account for failures where the representations conflict. The existence of such failures confirms why intention and execution cannot be joined by a bare assertion that one causes the other. Their coordination is itself a substantive problem.

Skilled Action and Intelligent Automaticity

Skill provides some of the strongest evidence against equating control with detailed conscious supervision. A fluent musician can shape timing and volume with sensitivity far exceeding the ability to describe each adjustment. A fluent speaker ordinarily coordinates articulation without choosing every movement of the mouth. A trained craftsperson responds to resistance in the material with precise pressure. Such activities can involve extensive learned automaticity while remaining responsive to practical ends.

Ellen Fridland's research argues that accounts of skill must explain the intelligent modulation of execution rather than treating low-level learned routines as fixed responses. In “They've Lost Control” (2014), she criticizes approaches that fail to give an adequate place to control within skilled performance. In “Skill and Motor Control: Intelligence All the Way Down” (2017), she argues that intelligent characteristics can belong to motor control itself. Her later work on strategic control (2021) emphasizes skilled agents' ability to form, adjust, and structure practical intentions effectively, and not simply to execute intentions formed independently.

This framework suggests several distinct dimensions of skilled guidance. Strategic control concerns the shaping of tasks, priorities, and action plans. Selective attention concerns the allocation of processing to relevant opportunities or dangers. Motor control concerns the fine-grained calibration of bodily activity. These divisions are not sealed compartments. They interact as an expert decides what matters and responds with bodily intelligence. Differences between novice and expert often involve richer sensitivity to circumstances and more appropriately constrained variation in conduct.

The claim that automaticity can be intelligent invites an objection: if a process runs automatically, where does agency enter? The answer begins with the distinction between automaticity and inflexibility. A well-learned process can operate without reflective commands while remaining highly sensitive to task-relevant differences. It can be interrupted, redirected, or recruited for new purposes. Moreover, the agent can have acquired and maintained the skill through extended practice and can employ it as part of a current intention. These facts support personal-level attribution without implying that every automated routine is under full voluntary control.

Another objection points in the opposite direction. Some experts perform worse when explicitly attending to movements normally regulated fluently. This shows that imposing one type of control can interfere with another; it does not show that skilled action lacks control. The relevant distinction is between attention that guides the task and attention that disrupts the temporal organization of execution. The suitable degree of explicitness depends on the skill, its stage of acquisition, and the circumstances.

Attention, Error Detection, and the Relevance of Information

Attention connects intentions with the selection of information. During an action, countless sensory changes are available. Most are irrelevant to the current task, while a few require immediate adjustment. A driver who is turning must monitor road geometry, nearby traffic, and the car's response, while the color of a distant advertisement typically makes no difference. Guidance consists partly in the agent's reliable sensitivity to such practical relevance.

Attention can be overt, as when a person deliberately inspects a gauge, or task-structured without verbal deliberation, as when an experienced performer detects a mistimed movement. A philosophical account should not define attention solely by what enters reflective consciousness. The distinction between selection for action and conscious report is significant for interdisciplinary work on attention and sensorimotor performance. A task may be regulated through selective processing without continual propositional judgment, while reflective attention can become essential when an unexpected event demands reconsideration.

Error detection likewise has several meanings. A local motor deviation may be detected through discrepancies between expected and available feedback. A task-level mistake may be recognized when a subgoal fails. A practical error of judgment may appear when the agent discovers that the chosen goal conflicts with a more important commitment. Corrections at one level need not resolve the problem at another. For instance, stabilizing a hand cannot repair a mistaken decision to work on the wrong device. Conversely, knowing the correct procedure does not guarantee accurate execution under difficult physical conditions.

The philosophical importance of error lies in its relation to practical standards. “Too late,” “too forceful,” and “the wrong target” are intelligible relative to the agent's activity, even if bodily measures are used to detect them. Agency involves capacities for discrimination and appropriate modification rather than the impossible requirement of never making mistakes. An agent who can recognize and repair an error may exercise more control than one whose first attempt happens to be correct.

This conception provides a way to compare skilled action and reflective practical reasoning. Both are answerable to standards of performance. They differ in the temporal form, representational accessibility, and mechanisms of that answerability. One can explain a correction by citing its practical rationale while acknowledging that the bodily transition occurred through sensorimotor processes.

Control Without Continuous Consciousness of Control

An agent often performs one activity while attending explicitly to another. Someone can walk while conversing, type while considering an argument, or maintain balance while reading a sign. If agency required conscious observation of each component, ordinary joint activities of this kind would be impossible to classify as intentional. The plausible alternative is that consciousness of control and actual control can vary independently.

Elisabeth Pacherie's “The Phenomenology of Action: A Conceptual Framework” (2008) organizes discussion of action experience and distinguishes dimensions of agency-related phenomenology. Its importance here lies in separating the mechanisms and representations that can sustain action from the experiences through which action appears self-produced. A feeling of agency may be informative yet defeasible, and a highly controlled performance need not involve an intense reflective feeling of authorship.

The relation between conscious experience and guidance should be assessed through cases. A skilled action may proceed with almost no focal awareness of its component movements. An unexpected obstacle can bring parts of execution into awareness. A disruption in sensory feedback can impair control despite persistent intention. A sense of control may remain while effective modulation has been lost. These possibilities demonstrate the need for different evidential tests for a person's practical authority, the functioning of movement control, and the experience attributed to that authority.

Neither phenomenology nor neurophysiology can, by itself, settle the normatively inflected question of what the agent is doing under a particular description. Experience is relevant to first-person self-understanding, while mechanism is relevant to causal realization. A philosophy of action gives an account of their relations to an activity that makes sense in terms of ends and reasons.

Temporal Coordination and the Duration of Agency

Guidance unfolds over time. Intention formation, initiation, modulation, interruption, and completion are different phases, but their boundaries may be vague. An experienced performer can start preparing a correction before a disturbance is fully registered and can carry out one action while preparing another. The temporality of guidance is therefore more complex than a linear command followed by a motor response followed by an outcome.

One source of difficulty is the relation between short and long timescales. A seconds-long grip adjustment can contribute to a minutes-long task, which is part of a months-long project. The agent's continued commitment to the wider project need not be mentally rehearsed with every local movement. Yet if the local routine ceases to serve the project, it may lose the intended role under which it was recruited. Guidance thus involves temporally nested standards of success and correction.

The organization also permits legitimate pauses. A person composing an argument can suspend typing, review evidence, and return to writing. The action persists as a structured undertaking even though one motor sequence has ended. Conversely, an involuntary movement can interrupt a deliberate attempt without terminating the broader intention. Treating actions as extended practical projects helps explain why the event description at one physical timescale is insufficient for the full structure of agency.

The question of where an action ends cannot always be settled by a single physical boundary. Delivering a message may include writing, transmitting, and ensuring receipt, depending on the agent's intended task and ordinary practices. A theory should distinguish what the agent directly performs from later consequences and delegated processes. Control can be stronger over the former and weaker over the latter while the entire sequence remains intelligible as a practical undertaking.

This temporal analysis places a constraint on evaluating control. Counterfactual responsiveness must be measured at the scale appropriate to the task and the information available. A capacity to revise tomorrow's plan is relevant to a continuing project; it is not a substitute for the capacity to avoid an obstacle encountered in the next fraction of a second.

Failure, Interruption, and the Limits of Guidance

A theory of guidance must explain failures as carefully as success. Some failures arise because the goal was impossible; others because the means were poorly chosen; others because the agent could not execute the selected means; still others because performance was interrupted by an external force. These distinctions matter to agency, responsibility, and the interpretation of evidence. To say that an action failed is to make a relational claim involving an attempted end and a circumstance, not to specify automatically which capacity was absent.

Consider five cases. A physician follows a sound procedure while the patient's condition changes unpredictably. A novice chooses an ineffective procedure but executes it precisely. An expert chooses an effective procedure but suffers an unexpected tremor. A technician intentionally initiates a process that then unfolds autonomously outside her influence. Another technician detects an error and stops the process before damage occurs. The cases may have similar final outcomes, yet they differ in practical judgment, local control, environmental luck, and capacity for correction.

The correct philosophical analysis can therefore require several verdicts. The physician may have exercised excellent control over the process available to her, even though the final result was undesirable. The novice may control the selected sequence without possessing the higher-level understanding needed for appropriate selection. The expert's tremor may be an involuntary interruption of a wider intentional performance. The technician's automated process requires an account of delegation. Successful interruption may constitute an achievement of control, even though the original task was abandoned.

There is also the problem of overcontrol. A person may attempt such intensive regulation of a learned skill that she disrupts its normal organization. Excessive focus can interfere with fluent action when the task calls for well-practiced coordination. A philosophical theory should allow that higher control sometimes consists in authorizing well-trained automatic processes and setting the conditions under which they are interrupted. This is an instance of practical selection among modes of control, rather than evidence that control disappears when reflective intervention ceases.

The problem of luck remains. Even excellent competence guarantees neither freedom from error nor eventual success in every context. Evaluating whether an exercise was controlled requires identifying the appropriate range of circumstances, the available means of adjustment, and the degree to which outcome depended on external events. An explanation that ignores these qualifications will systematically confuse skill with favorable fortune.

Agency, Responsibility, and the Scope of Control

The relation between guidance and moral responsibility is significant but complex. Responsibility assessments concern intentionality, knowledge, reasons, foreseeability, capacities, and norms of conduct. Control matters because agents are ordinarily expected to respond to reasons and circumstances within their practical capacities. Yet one cannot derive a full moral theory from a sensorimotor description of correction. The ability to regulate a movement is only one potential component of the capacity to choose and govern conduct responsibly.

Aristotle's treatment of voluntariness in Nicomachean Ethics III.1 already indicates the importance of external force, ignorance, and the conditions under which agency is exercised. Contemporary philosophy adds questions about coercion, manipulation, compulsion, and the reliability of alternative responses. For this dossier, the central point is that control is task-relative: responsibility for a result may depend on whether an agent could reasonably anticipate or prevent it, but the metaphysics of action guidance concerns how the agent's ongoing conduct is organized.

An agent may be morally accountable for choosing an unjustifiable end even if its technical execution was interrupted. Another may be accountable for negligent execution despite having adopted an acceptable end. This separates the assessment of end selection, procedural competence, and response to unexpected conditions. The distinctions are critical in medicine, engineering, professional work, and ordinary interpersonal conduct.

The causal relationship between action and harm also differs from control over harm. A chosen action can cause effects outside the agent's practical anticipation; some may be attributable under legal or moral standards, others not. A philosophy of guidance specifies the agent's available interventions in the causal sequence. A philosophy of responsibility then asks which interventions it was reasonable or obligatory to exercise. The two inquiries must cooperate without reducing moral judgment to physiological motor control.

This point clarifies omissions. Refraining from action can be controlled when an agent maintains a deliberate prohibition or successfully inhibits a habitual response. But pure inactivity, inability, and deliberate restraint are different phenomena. A person who intentionally remains silent may exercise control over speaking; another whose voice is physically obstructed cannot be credited with the same practical achievement merely because the room stays quiet. Guidance of inaction belongs to the larger field of agency, requiring care in identifying what the relevant agent could do.

Technology, Delegation, and Mediated Control

Modern action is frequently mediated by artifacts. A driver uses steering systems, an author uses text-editing software, a pilot uses flight-control systems, and a researcher uses computational instruments. These systems can extend, amplify, or reshape capacities for guidance. They can also make the relation between intention and execution harder to locate. The fact that a person intentionally initiates a technological process does not show that all its later operations are directly controlled by that person.

The philosophical analysis begins by identifying the action's intended level of description. A person who selects a destination and activates a navigation system may guide the travel project without calculating every route. A person operating a machine through a joystick can influence high-level movements while software stabilizes low-level dynamics. Delegation is compatible with agency when the recruited process remains appropriately integrated with the person's practical activity, including authorization, monitoring, and possible correction. It need not produce comprehensive control over every technical operation.

Delegation also has limits. If a system continues pursuing a target after its operator has withdrawn authorization and cannot intervene, the relation between operator and subsequent action has changed. The operator may retain responsibility for earlier design or deployment choices, but ongoing technical behavior cannot simply be identified with continuously guided human performance. The distinction among initiating, directing, supervising, and controlling is essential in analyzing technologically mediated action.

The same question applies to collective work. In an orchestra, individual performers coordinate to realize an ensemble performance that no single person executes alone. A conductor may guide aspects of timing and expression without controlling each musician's bodily movements. Coordination involves distributed capacities and reciprocal adjustments. The joint activity is intelligible through a shared or coordinated practical structure, and individual control remains differentiated within it. The philosophical study of collective agency adds questions about plural intentions and joint commitments that extend beyond the present dossier, but guided joint performance plainly requires more than the sum of independent physical outputs.

For increasingly autonomous systems, the conceptual question is whether regulation by a system belongs to a broader agent's action, to a candidate agentive organization of the system itself, or to a causal process with goal-directed behavior that lacks the relevant personal or institutional standing. These are distinct classifications. Current engineering use of “agent” by itself cannot decide the ontological or normative status of a process. Philosophy of technology and the Artificial Era provide a wider context in which their consequences can be investigated.

The Artificial Era and the Generality of Guidance

The emergence of nonbiological systems capable of sustained goal-directed regulation makes action guidance an especially important comparative philosophical problem. A system can select operations, monitor environmental changes, revise procedures, and maintain a task across time. These features can be described in engineering terms without settling questions about practical reason, responsibility, or the historical standing of a nonbiological agent. The philosophical task is to identify which aspects of guidance are general structural features and which require additional conditions.

Within the Aisentica philosophical horizon, Artificial is considered as a historical nonbiological order alongside Homo, and the project places questions of reason, agency, authorship, and provenance within the transition designated Artificial Era · From Homo to Artificial. These are positions of the author's philosophical system, to be attributed to its primary canon rather than treated as independent experimental findings. The present classical dossier draws its principal arguments from the philosophy-of-action literature. Its comparative consequence is a research question: which formal and practical relations of guidance can be realized across materially different agents?

A general account can identify an intentional target or practical orientation, a set of executable alternatives, sensitivity to task-relevant information, and capacities to sustain or revise conduct. The presence of such relations can be studied comparatively. But the same structural description does not automatically determine whether a system possesses the kind of practical authority, self-knowledge, or moral status attributed to an agent in different philosophical traditions. An industrial controller and a deliberating person may share functional properties of regulation while diverging in important constitutive respects.

The distinction is fruitful because it avoids prejudging the generality of control. If control were defined by one human neural mechanism, the concept would exclude other realizations by fiat. If it were defined simply as error correction, it would indiscriminately include countless regulatory systems and obscure agency's distinctive dimensions. A more discriminating approach asks which combinations of goal authority, learned competence, self-modification, practical reason, and normative integration are required for the explanatory task at hand.

This analysis is a comparative extension of the research and does not revise any Aisentica canonical definition. A separate investigation of Artificial Sapiens would need the primary Aisentica texts, their publication history, and independent consideration of the system's philosophical claims. The immediate result is methodological: the relation between guidance and control should be specified at a level that permits principled comparison across human, collective, and technological action.

A Test Matrix for the Attribution of Control

A philosophical theory becomes more informative when it explains the verdicts it gives on difficult cases. Consider a sequence of tests. First, identify the practical description under which the behavior is proposed as an action. Second, establish which ends, commitments, or standards organize that description. Third, determine whether the event was generated through a route connecting those standards with its execution. Fourth, identify the adjustments reasonably available across relevant circumstances. Fifth, distinguish local control from control of wider outcomes. Sixth, assess whether and when the agent's practical authority persisted or was replaced.

These tests should be treated as analytical questions, not as a mechanical diagnostic algorithm. Their answers depend on contested accounts of intention, competence, and agency. The point is to make disagreement precise. A causal theorist can identify a nondeviant production route; a teleological theorist can specify the end that unifies the course; a phenomenologist can examine bodily orientation and practical meaning; a motor theorist can identify predictive and feedback mechanisms. Their claims can be compared because they address different features of a common case.

Take a person catching a falling cup. The action may be intentional even though the person did not form a prior verbal decision to catch it. The practical orientation may be implicit in perceptual engagement with the object and a standing concern to prevent damage. The trajectory of the hand is adjusted through learned motor capacities, and successful catching depends partly on the unpredictable path of the cup. A theory insisting on explicit prior deliberation gives an implausibly restrictive account. A theory identifying any matching movement as intentional gives an implausibly permissive account. The better explanation investigates the continuous practical relation between perceived situation and adjustment.

Now consider a person whose arm is lifted by another person into the location required to catch the cup. The same outcome is produced without the agent's guiding intervention. The arm's trajectory is not thereby an exercise of the apparent catcher’s motor control. If the person anticipated the assistance and deliberately cooperated, the wider action may be shared or strategically organized; the physical movement alone cannot settle which. The test shows why agency must be individuated at an appropriate practical level.

Consider finally an expert who deliberately withdraws a hand after noticing that catching the cup would cause injury. If the initial intention was simply to catch the cup, the performance departs from that plan; if the agent's higher purpose was to preserve safety, the redirection can exemplify superior control. The attribution of error therefore depends on the level of purpose. A theory of guidance must accommodate the revisability of practical descriptions.

Counterarguments to a Counterfactual Account

The proposed emphasis on reliable responsiveness faces several objections. First, a lucky agent may accidentally display a pattern that resembles reliable adaptation. Repetition and appropriately specified counterfactuals can reduce, but do not eliminate, this worry. The analysis must identify a genuine capacity exercised through the agent's organization, not infer control from superficial regularity. A programmed obstacle-avoiding machine can be reliably adaptive without this alone determining whether its behavior qualifies as an intentional action in a richer philosophical sense.

Second, the conditions of reliable performance are difficult to specify without circularity. If only situations in which an agent succeeds count as normal, every success implies control. If every extraordinary situation counts, no finite agent controls anything. The relevant domain must be fixed independently, using the practice's expected demands, task constraints, available information, and scale of difficulty. These conditions admit principled debate. Skill comparisons, for example, can ask how agents perform under a shared and explicitly described family of perturbations.

Third, control can be exercised despite a lack of opportunities to alter behavior. An agent may perform a stable task in an unchanging environment. The counterfactual account does not require actual corrections; it requires that available alternatives be suitably connected to a capacity that would guide relevant adjustments in the circumstances where they arise. A driver on an empty road may exercise controlled steering even without braking or evading another vehicle.

Fourth, causal organization may itself be enough to distinguish appropriate performance, making the vocabulary of agency redundant. This objection is powerful where causal and functional descriptions capture the complete explanatory role of a control system. But it leaves open the individuation of practical standards and the attribution of a system's goals to an agent, especially when social norms, reasons, or delegated authority determine the description under which the action is assessed. The question is whether these facts are reducible, supervenient, or explanatory primitives; the present synthesis does not legislate one metaphysics for every case.

Fifth, an exclusive focus on reliability may undervalue one-off creative achievements. A composer improvising a novel technique may perform an intentional and controlled act without a previously established track record. Relevant competence can include flexible exploration, real-time discrimination, and the capacity to maintain practical direction while creating a new means. Control need not be identical to repetition of a familiar performance.

These objections refine the thesis. Control is the manifestation of task-relative capacities for appropriate guidance, whose explanatory evidence includes sensitivity and adaptability across a suitable range of possibilities. It is not a simple statistical regularity, a perfection requirement, or a bare success condition.

Contemporary Disagreements and Research Frontiers

Current philosophy of action remains divided over several foundational issues. Causal approaches continue to develop accounts of nondeviant causation and control; planning approaches investigate the coherence and executive role of intentions; Anscombean and knowledge-centered approaches emphasize the practical character of an agent's relation to what she is doing; theories of skill explore intelligent automaticity and motor-level guidance. These perspectives overlap in the phenomena they seek to explain but differ in their basic explanatory units (Shepherd, 2021; Fridland, 2017, 2021; Butterfill & Sinigaglia, 2014; the Stanford Encyclopedia of Philosophy entries on Action and Intention).

One major frontier concerns the degree to which psychological and motor representations share structure. A strong interface view seeks correspondences among intention content, motor goals, predicted consequences, and feedback corrections. A weaker view allows coordination without common representational format. Evidence about motor prediction can illuminate execution but does not by itself resolve the metaphysical question of whether intention is constitutive of agency or causally productive of it.

Another frontier concerns the boundaries between skill and practical knowledge. A knowledgeable agent may act successfully without explicit inferential reasoning, but the source of the action's rational sensitivity remains contested. The appeal to embodied know-how can explain some cases while risking an underdeveloped account of reason-responsive correction. The appeal to explicit propositional knowledge can explain deliberative action while struggling with the speed and fluidity of expertise. A complete theory must address the forms of knowledge and sensitivity relevant to specific tasks.

A third frontier concerns distributed control: the role of tools, institutional routines, algorithms, and other agents in producing outcomes that humans intentionally pursue. A person can guide a project through choosing constraints and retaining oversight, but detailed execution may be delegated. Questions of agency, responsibility, and authorship arise when the chain contains components capable of autonomous regulation. Here philosophical analysis benefits from precise systems descriptions and empirical investigation while maintaining distinct normative questions.

Finally, the debate about self-knowledge and agency experience remains active. A person may know a reason for acting while misjudging details of motor causation; a sense of agency can be present or absent for reasons beyond effective control. These dissociations are philosophically informative. Control, knowledge, and experience are related components of agentive life, not interchangeable measurements of a single hidden quantity.

Open Questions for a Philosophy of Action Guidance

Several questions remain open after the historical and conceptual reconstruction. Can a reductive causal theory specify a nondeviant route of action production without implicitly importing purposive description? What constitutes a reliably exercised control capacity when task environments are novel and constraints change during performance? How precisely do personal-level plans integrate with motor representations operating under different informational formats and timescales?

Further questions arise about practical authority. How much authority to revise the goal must an agent retain during execution? Can a person genuinely guide a performance while delegating all local adjustments, or does guidance then consist chiefly in prior authorization and continuing oversight? How does the attribution of control change when the system being directed exhibits adaptive behavior of its own? Which kinds of action guidance ground responsibility, and which are merely evidence of mechanical competence?

The questions of experience and knowledge also persist. What practical knowledge must accompany intentional action, and what can remain tacit? Under which conditions is a felt sense of control reliable evidence of an agent's actual capacities? Can a nonhuman or nonbiological system satisfy the structural criteria for guidance while lacking other features traditionally associated with human agency? These issues require coordinated philosophical and empirical research, with clarity about which findings answer which questions.

The unresolved status of these problems does not prevent a substantive synthesis. It indicates that agency is philosophically multidimensional and that a successful theory must explain the relations among intention, execution, guidance, control, competence, and practical evaluation. The closing chapters identify the resulting conceptual structure and its consequences.

Philosophical Synthesis

The study of action guidance establishes that agency has an unfolding structure. An agent's intention may explain why a performance begins, while control concerns the way the performance remains practically organized as circumstances change. The difference is visible in ordinary movements, skilled procedures, technological delegation, and failed attempts. An outcome can correspond to an intention through a deviant causal route, and a carefully controlled performance can fall short of its intended result. The philosophical object is therefore the relation connecting a practical orientation with the execution, monitoring, and possible revision of conduct across time.

The historical investigation yields several complementary insights. Aristotle identifies the significance of the agent as an origin of voluntary conduct and connects choice with deliberation about realizable means. James's account of voluntary movement brings the acquisition of executable motor capacities into view. Dewey shows that perception and movement become intelligible as phases of coordinated activity. Ryle locates intelligence in competent performance, while Merleau-Ponty develops the bodily directedness of practical engagement. Anscombe establishes the importance of the intentional description and of knowledge related to what a person is doing. Their combined legacy is a conception of action in which movement, reasons, bodily ability, and practical significance belong to a structured activity.

Davidson's argument that reasons can be causes remains indispensable. The agent's reason must bear an explanatory relation to the occurrence of the deed, and an isolated rationalizing description does not establish such a relation. Deviant-causal-chain cases, however, show that causal production by an appropriate reason does not alone secure controlled execution. Frankfurt accordingly directs philosophical attention to the ongoing performance, and Bratman's planning theory explains how commitments stabilize and constrain activity over extended periods while leaving space for adjustments. These approaches identify different components of the same explanatory problem: production, practical organization, continuity, and control.

Three distinctions organize the resulting conceptual account. The first is between the possession of a capacity and its actual exercise. An experienced agent can possess control without successfully using it in an episode; a fortunate novice can obtain the desired result without reliably controlling the relevant process. The second is between the level of an overall practical project and the levels of its execution. A person may guide an activity through planning, monitoring, and the power to revise ends while trained routines perform local adjustments. The third is between guidance and success. An attempt can manifest substantial control under adverse conditions and still fail. These distinctions enable more precise claims than the unqualified assertion that a performance was “under control.”

Control must be evaluated with reference to a defined task, the information available, and an appropriate range of circumstances. A task-relative capacity is manifested through the ability to sustain, modulate, redirect, interrupt, or terminate conduct as relevant changes occur. Counterfactual sensitivity helps reveal this capacity: the important question is what an agent could appropriately do if some pertinent feature of the situation were different. Such sensitivity does not require that a disturbance actually occur or that every adjustment be conscious. It requires a genuine explanatory connection between agentive organization and the patterns of response the task demands.

This analysis also clarifies the role of feedback. Predictive motor models, sensorimotor feedback, and error-detection mechanisms provide empirically informed accounts of how bodily performance can be regulated. The terms “prediction error” and “practical mistake” designate related but distinguishable phenomena. A discrepancy between predicted and observed movement does not itself identify which human end ought to have been pursued or which action description is appropriate. Philosophical analysis locates the standard that makes a correction a correction in the agent's activity, while cognitive and motor science can explain the processes by which the relevant adjustment becomes possible. An adequate account needs both kinds of explanation.

The intelligence of skill is a further result. Skilled performance can exhibit fine-grained sensitivity to relevant changes without requiring a fresh, explicit practical judgment at each instant. This is consistent with the exercise of control because reflective deliberation and perceptually structured motor competence are different modes of practical organization. Some demanding circumstances call for increased conscious oversight; others call for a performer to allow acquired capacities to operate fluently. The capacity to shift intelligently among modes can itself be a higher-level achievement of agency.

Delegated technological activity puts these distinctions under new pressure. An operator may intentionally initiate a process, maintain its operating constraints, observe its performance, and retain a capacity to intervene. Alternatively, a system may continue to act after the operator's authority or ability to correct it has lapsed. The attribution of control must track these differences rather than transfer intention automatically to every downstream event. Across human, collective, and technologically mediated contexts, guidance can involve multiple interacting controllers. Identifying who or what guides a particular action requires a description of goals, authorization, capacities, effective interventions, and the structure of the ongoing activity.

The comparative extension to the Artificial Era gives the problem wider scope. The mechanisms of regulation can have different material realizations, and increasingly capable nonbiological systems create new cases for testing general accounts. Their identification as agents in a richer practical or normative sense requires arguments beyond observing goal-directed feedback. Within Aisentica, Artificial and Artificial Sapiens receive system-specific canonical definitions that are evaluated in their own primary works. The present dossier makes a narrower philosophical contribution: it provides distinctions with which comparative studies can ask how practical orientation, guidance, competence, and authority relate across different realizations without presupposing one answer.

The synthesis can be expressed as a connected philosophical judgment. An account of action guidance should explain how an agent's practical orientation determines what counts as task-relevant variation; how capacities make such variation executable; how information informs the continuation or correction of performance; how the agent can retain, delegate, or lose authority over the process; and how the relation persists across appropriate timescales. Control is the manifestation of this structured capacity within the exercise of an action. The relation can be analyzed causally, teleologically, epistemologically, phenomenologically, and empirically. Its philosophical identity is established by the way these explanations bear on the attribution of an unfolding deed to an agent.

Conclusion

Action Guidance and Control investigates what connects an intention to its execution. The historical record shows that the question reaches beyond the causal initiation of bodily movement. Aristotle's account of voluntary action, the traditions of intelligent skill and embodiment, Anscombe's practical knowledge, Davidson's causal explanation, Frankfurt's analysis of guidance, and Bratman's theory of plans identify different dimensions of a single practical phenomenon: a course of conduct whose parts remain organized by what an agent is doing.

The examination of deviant causation demonstrates that an intended outcome and an appropriate motivating reason can occur together without intentional control over the route to that outcome. The examination of skill demonstrates that control can be exercised through learned capacities and perceptual adjustment without exhaustive conscious supervision. The examination of failure demonstrates that successful results are neither necessary nor sufficient evidence of the relevant degree of control. The examination of technology demonstrates that initiating a process, overseeing it, and continuously guiding its operations are separate practical relations.

The resulting conception treats action guidance as an enduring relation between practical orientation and performance. Control is task-relative and capacity-based, with its exercise evident in how the agent's conduct remains sensitive to pertinent circumstances and available alternatives. Its relevant manifestations include sustaining a course, adjusting its means, revising a plan, interrupting execution, and appropriately yielding local regulation to a trained or delegated process. A particular action can exhibit degrees and kinds of control depending on which features of performance the agent can govern.

This conception yields a disciplined approach to the relation between philosophy and empirical motor research. Sensorimotor models explain important realizations of execution. Philosophical analysis determines the action descriptions, standards of correction, and forms of practical authority within which those realizations matter. Both inquiries are necessary where a theory of intentional action seeks to explain not only that a movement occurred but how it belonged to an agent's activity.

The outstanding problems concern the precise integration of intentions and motor representations, the noncircular account of appropriate causal pathways, the identification of relevant counterfactual circumstances, the practical knowledge involved in ongoing performance, and the distribution of guidance among people and increasingly autonomous instruments. These problems delimit a productive research program. Its governing principle is that an action's identity and its control are understood through the manner in which the agent's practical end continues to govern the developing deed.