Moral Responsibility

For most of the contemporary debate, moral responsibility is the real prize. The metaphysics of determinism and free will matters because of what it might do to responsibility — to the fairness of praise and blame, reward and punishment, guilt and indignation. This page steps back from the positions and examines responsibility itself: what it is, what its different faces are, what conditions an agent must meet to bear it, and how tightly it is really tied to free will. The payoff is a reframing that organizes everything else: the sharpest modern question is often not "do we have free will?" but "what kind of responsibility can creatures like us actually have?"

This is the technical companion to the plain-language treatment in What Is Free Will?; it develops the backward-looking / forward-looking distinction sketched there into the full taxonomy.

Free will and moral responsibility: two questions, not one

It is tempting to treat "free will" and "moral responsibility" as a single topic, but they are distinct, and prying them apart is one of the most important moves in the modern literature.

• Free will is standardly a claim about a capacity — typically the power of self-determination, or the ability to do otherwise, or ultimate sourcehood.
• Moral responsibility is a status: being an apt target of a distinctive family of responses (praise, blame, the reactive attitudes) and the practices built on them (desert, punishment, answerability).

The two come apart because one can ask whether responsibility survives determinism without first settling what free will requires. Semicompatibilism (Fischer) is precisely the view that moral responsibility is compatible with determinism even if the freedom to do otherwise is not — so it brackets the free-will question and argues about responsibility directly. Free will is often defined as "the strongest control condition on moral responsibility," which makes responsibility the more fundamental notion and free will its servant.

Holding responsible vs. being responsible

A second crucial distinction. There is:

• being responsible — a fact about the agent (she really is an apt target of blame), and
• holding responsible — an activity of others (we blame her, resent her, demand an account).

Which is prior? The realist takes being responsible as basic and holding responsible as a response that is correct when the fact obtains. P. F. Strawson's great reversal (in "Freedom and Resentment," 1962) was to suggest the practice of holding responsible comes first: there is no fact of responsibility floating free of our responses; to be responsible just is to be a fit object of the reactive attitudes. This makes the theory of responsibility partly a theory of when those attitudes are appropriate — and removes the need for a prior metaphysical proof that we are "really" free.

The reactive attitudes (Strawson)

Strawson's starting point is a set of emotions through which we register the quality of will others bear toward us: resentment, indignation, gratitude, guilt, forgiveness, hurt feelings, certain kinds of love. These reactive attitudes are the natural expression of our investment in being treated with goodwill rather than malice or indifference. To hold someone responsible is to be susceptible to these attitudes in response to what their conduct expresses.

Strawson contrasts two stances we can take toward a person:

• The participant (or reactive) stance — engaging with them as a fellow member of the moral community, open to resentment and gratitude.
• The objective stance — viewing them as an object to be managed, treated, or trained, as we might a hazard, a small child, or someone in the grip of psychosis. Here the reactive attitudes are suspended.

When do we suspend them? Strawson distinguishes two kinds of pleas:

• Excuses ("he didn't mean to," "it was an accident," "he was pushed") — these concede the agent is a responsible agent but deny the particular act expressed ill will. The act is reinterpreted; the agent remains within the moral community.
• Exemptions ("he's a small child," "he's psychotic," "he was under hypnosis") — these withdraw the agent as such from the participant stance, classifying them as not (now, or in this respect) a fit target of the reactive attitudes at all.

Strawson's anti-skeptical thesis is that the general truth of determinism is neither an excuse (it does not reinterpret any particular act) nor a global exemption (we could not, and psychologically cannot, take the objective stance toward everyone permanently). So responsibility is not hostage to the metaphysics.

The faces of moral responsibility

Strawson's account was later refined by the recognition that "responsibility" names several distinct things, which can come apart. The influential taxonomy runs:

• Attributability (Gary Watson, "Two Faces of Responsibility," 1996) is the weakest, most agent-disclosing sense: an act is attributable when it is an expression of the agent's real self — her deep self, values, or quality of will. It grounds judgments of virtue and vice but not yet of liability. (This is the home of the deep-self / real-self compatibilist views.)
• Answerability (T. M. Scanlon, Angela Smith, Pamela Hieronymi) makes the criterion rational: you are responsible for those of your attitudes and actions that are judgment-sensitive — connected to your assessment of reasons — because you can, in principle, be asked to defend them. On this view even spontaneous attitudes (a flash of contempt) are responsibility-apt if they reflect your evaluative take on the world.
• Accountability (Watson; R. Jay Wallace; Stephen Darwall) is the strongest, most demanding sense — the one closest to Strawson's blame and to desert. To hold someone accountable is to subject them to a legitimate demand backed by the reactive attitudes and liability to sanction. This face carries the heaviest conditions, because holding someone to account (unlike merely appraising their character) must be fair.

David Shoemaker (Responsibility from the Margins, 2015) develops the tripartite view that these are three genuinely different kinds of responsibility, each with its own conditions, so that an agent can be responsible in one sense but not another (a psychopath may be attributability-responsible — his cruelty is truly his — yet not fully accountability-responsible).

Two theories of what "holding accountable" amounts to

Within the accountability face, two influential accounts say what blame is:

• The normative-expectation / fairness account (R. Jay Wallace, Responsibility and the Moral Sentiments, 1994). To hold someone responsible is to subject them to reactive attitudes governed by normative expectations (the moral demands we legitimately make). The central question becomes one of fairness: it is appropriate to blame someone only if it would be fair to demand that they have complied — which shifts the debate from "could they have done otherwise?" to "is the demand fair given their capacities?"
• The conversational account (Michael McKenna, Conversation and Responsibility, 2012). Responsibility is a conversational exchange of meaning: a wrong act is a moral contribution ("I may treat you this way"), blame is a reply ("no, you may not"), and the agent can respond in turn (apology, justification, excuse). Stephen Darwall's second-personal account is a cousin: blame is an addressed demand that presupposes the other as an accountable party who can recognize and answer it.

Basic desert vs. forward-looking responsibility

The deepest fault line — and the hinge of the free-will-skepticism debate — concerns why a responsible agent may be blamed or punished.

• Basic-desert responsibility (the focus of Derk Pereboom and the skeptics). An agent is responsible in the basic-desert sense if she deserves blame or praise non-instrumentally and non-contractually — simply in virtue of having knowingly done wrong (or right), as a matter of what she fundamentally merits, independent of any further good the blame might do. This is the strong notion that grounds retributive punishment and the harshest reactive attitudes, and it is exactly what the hard incompatibilist argues no one can have.
• Forward-looking (consequentialist or relational) responsibility. Holding responsible is justified by what it achieves or by the relationships it sustains — moral formation, reconciliation, protection, the integrity of our practices of mutual address — rather than by desert. This sense may survive even if basic desert does not, which is why skeptics can keep a reformed "responsibility" while rejecting retribution.

Other non-basic justifications sit between: consequentialist (Schlick, Smart — blame as a tool that "shapes the springs of action"), contractualist/fairness (Scanlon, Wallace), and Strawsonian (the attitudes are constitutive of personal relationships and need no external vindication). The whole punishment debate is downstream of this distinction.

The conditions on responsibility

Since Aristotle (Nicomachean Ethics III), two conditions have been thought jointly necessary for responsibility — and they map neatly onto Strawson's two pleas:

1. The control (or freedom) condition. The action must be appropriately up to the agent — under her control, voluntary, not compelled or coerced. This is the condition free will is meant to satisfy, and the whole compatibilism/incompatibilism dispute is a dispute about how strong it must be (mere reasons-responsive guidance control, or full leeway/sourcehood?). Failures here are exemptions (no capacity) or excuses of compulsion/coercion.
2. The epistemic (or knowledge) condition. The agent must be aware, in the right way, of what she is doing and of its moral significance — its nature, consequences, and wrongness. Failures here are excuses of ignorance ("she didn't know, and couldn't reasonably have known, the glass was poisoned"). A large literature debates whether the relevant ignorance must be culpable (traceable to an earlier akratic or negligent failure) to preserve responsibility.

Tracing. Both conditions can be satisfied derivatively. A drunk driver may lack control and knowledge at the moment of the crash, yet be fully responsible because the incapacity traces back to an earlier exercise of control and knowledge (the decision to drink and drive). Fischer–Ravizza and others use tracing to handle responsibility for the unforeseen, the automatic, and the self-induced.

Does moral responsibility require free will — or alternatives?

This is where the responsibility debate loops back to the metaphysics, and the key question is whether responsibility needs the ability to do otherwise (the leeway condition) or only the right kind of actual-sequence source (the sourcehood condition).

• The Principle of Alternative Possibilities (PAP): an agent is morally responsible for what she did only if she could have done otherwise. If PAP is true, then any argument that determinism removes alternatives (the Consequence Argument) directly threatens responsibility.
• Frankfurt's challenge. Frankfurt-style cases purport to show PAP is false: an agent can be responsible even when a counterfactual intervener would have prevented any alternative, because she in fact acted on her own. If this succeeds, responsibility is decoupled from leeway and rests on the actual sequence — on whether the action issued from the agent in the right way.
• Guidance control and semicompatibilism. Fischer–Ravizza's answer: responsibility requires guidance control — action from the agent's own, moderately reasons-responsive mechanism — not regulative control (access to alternatives). Because guidance control is an actual-sequence, dispositional notion, it is compatible with determinism. Hence semicompatibilism: responsibility can be safe even if the Consequence Argument shows the freedom-to-do-otherwise is not.
• Source incompatibilist dissent. Source incompatibilists grant Frankfurt's point about alternatives but insist responsibility still fails under determinism, because the actual-sequence source is itself a link in a chain the agent did not originate — the manipulation arguments press exactly this.

So the structure of the whole field can be read off the responsibility question: leeway theorists keep PAP and fight over the Consequence Argument; source theorists drop PAP and fight over manipulation and ultimacy; compatibilists of the guidance-control stripe drop PAP and locate responsibility in reasons-responsiveness.

Common misconceptions

• "Responsible just means caused by the agent." Mere causal authorship is not enough; responsibility adds the control and epistemic conditions, and (on most views) fitness for the reactive attitudes. A falling reflex or a coerced act may be "caused by you" yet not something you are responsible for.
• "Moral responsibility stands or falls with free will." Not quite: semicompatibilism defends responsibility while conceding the freedom-to-do-otherwise may be lost, and the different faces have different conditions, so responsibility can partly survive even where one strong sense lapses.
• "Blame is just a feeling / just a useful tool." These are two theories (Strawsonian sentimentalism, consequentialism), not the default. Desert-based and answerability accounts hold that blame answers to facts about the agent and reasons, not merely to utility or emotion.
• "To be responsible you must deserve to suffer." That is only basic-desert accountability — the strongest face. Attributability (this vice is yours) and answerability (defend your judgment) are genuine forms of responsibility that carry no commitment to deserved suffering.

Where moral responsibility sits

Moral responsibility is the destination the rest of the free-will debate is traveling toward. Pulling it apart into attributability, answerability, and accountability, and separating basic desert from forward-looking justifications, does real work: it shows that the blunt question "are we responsible?" hides several questions with different answers, and that a position can give up the most demanding sense (basic-desert accountability) while keeping the rest. That is exactly the space the hard incompatibilist exploits (drop basic desert, keep forward-looking accountability), that the compatibilist occupies (responsibility needs only guidance control and fitness for the reactive attitudes), and that the libertarian resists (only ultimate sourcehood secures genuine desert). Whichever way one leans, the lesson is that "free will" is best understood as shorthand for "whatever control moral responsibility requires" — so getting clear about responsibility is getting clear about what was at stake all along.