Semi-Compatibilism: What If Free Will and Moral Responsibility Come Apart?

Imagine two people commit the exact same crime. Both of them steal a car. One of them could have chosen not to steal it. The other couldn't have done otherwise; some configuration of the universe made this action inevitable. Here's the strange question: are they equally responsible?

Most of us feel like they are. The guy who steals a car because he decided to seems just as guilty as the guy who steals a car because he decided to, even if some philosopher informs us that one of them "couldn't have done otherwise." After all, both of them wanted to steal the car, both of them deliberated (however briefly), and both of them acted on their own reasons. The fact that one of them was, unbeknownst to anyone, causally determined to act that way doesn't seem to change the moral picture one bit.

What if that intuition is right? What if moral responsibility has nothing to do with whether you could have done otherwise, and everything to do with how you actually made your decision?

That is the core thesis of semi-compatibilism, and it is one of the most creative and strategically brilliant moves in the entire free will debate. In this post, I want to lay out exactly how the argument works, where it draws its strength from, and where the most serious objections land.

Where We Are in the Debate

Over the last two episodes, we've been wrestling with the Principle of Alternative Possibilities (PAP), the idea that moral responsibility requires the ability to do otherwise. Frankfurt cases challenge PAP by constructing scenarios in which an agent seems morally responsible even though she couldn't have acted differently. The dilemma defense pushes back, arguing that no coherent Frankfurt case actually eliminates genuine alternatives. And general-abilities accounts offer yet another line of defense for PAP, distinguishing between specific and general capacities.

But now we encounter a philosopher who makes a different kind of move entirely. John Martin Fischer doesn't try to defend Frankfurt cases against the dilemma defense, and he doesn't try to show that determinism is compatible with alternative possibilities. Instead, he does something more radical. He concedes that the incompatibilists might be right that determinism rules out the ability to do otherwise. And then he says: so what?

The Big Idea: Splitting the Package

Most positions in the free will debate treat free will and moral responsibility as a package deal. If you have free will, you're morally responsible. If you lack free will, you're not. The two concepts travel together like mass and weight.

Fischer's insight, developed across a series of publications beginning in the 1980s and given its fullest treatment in Responsibility and Control: A Theory of Moral Responsibility (1998), co-authored with Mark Ravizza, is that this package deal is an assumption, not an established truth. What if free will and moral responsibility are actually separable? What if you could lack free will in the libertarian sense (the ability to do otherwise, given the exact same past and laws) and yet still be morally responsible for your actions?

Fischer calls his position semi-compatibilism because it is compatibilist about one thing (moral responsibility) but potentially incompatibilist about another (free will). He's happy to grant, at least for the sake of argument, that determinism is incompatible with the ability to do otherwise. What he denies is that this matters for moral responsibility.

This is a strategic masterstroke. The Consequence Argument, formulated by Peter van Inwagen in An Essay on Free Will(1983), is one of the most powerful arguments in all of philosophy. It purports to show that if determinism is true, then no one has ever been able to do otherwise than what they actually did, because our actions are the inevitable consequences of the past and the laws of nature, neither of which is up to us. Fighting this argument head-on has proven enormously difficult. Fischer's move is to sidestep it entirely. You want to say determinism rules out alternative possibilities? Fine. Moral responsibility doesn't need them.

The core argument can be stated formally:

(1) Moral responsibility requires a certain kind of control over one's actions.

(2) The relevant kind of control is guidance control, not regulative control.

(3) Guidance control does not require the ability to do otherwise.

(4) Therefore, moral responsibility does not require the ability to do otherwise.

(5) Therefore, even if determinism rules out the ability to do otherwise, it does not rule out moral responsibility.

Everything hangs on the distinction between regulative control and guidance control, and on whether guidance control really is sufficient for moral responsibility. So let's look at that distinction carefully.

Two Kinds of Control

Fischer distinguishes between two fundamentally different kinds of control an agent can have over her actions.

Regulative control is the ability to determine what happens, combined with the ability to have done otherwise. Think of a driver on an open road with a working steering wheel. She can turn left, she can turn right, she can go straight. She has control over which outcome obtains, and she has access to alternatives. Regulative control is what libertarians typically have in mind when they talk about free will.

Guidance control is the ability to guide your action in the way you actually do, without necessarily having access to alternatives. Think of a driver whose steering wheel happens to be locked in position, though she doesn't know it. She drives straight ahead because she wants to drive straight ahead, using her own practical reasoning. If she had wanted to turn, she couldn't have. But she didn't want to turn. Her actual action was guided by her own reasons, her own deliberation, her own mechanism of decision-making.

Fischer's claim is that moral responsibility requires only guidance control. And guidance control, crucially, is compatible with determinism, because it concerns only the actual sequence of events (how you actually arrived at your action), not any alternative sequence (what you could have done instead).

This tracks a distinction Fischer borrows from the philosophical literature between what he calls "actual-sequence" reasons and "alternative-sequence" reasons. The actual-sequence approach asks: was the process by which you arrived at your action the right kind of process? The alternative-sequence approach asks: could you have done something different? Fischer bets everything on actual-sequence evaluation.

Moderate Reasons-Responsiveness

If guidance control is going to bear the full weight of moral responsibility, Fischer needs to say precisely what it consists in. His answer, developed in detail with Ravizza, is that guidance control requires two things: (a) that the agent's action flow from a moderately reasons-responsive mechanism, and (b) that the agent own that mechanism. Let's take these in order.

A mechanism is reasons-responsive when it is sensitive to reasons for and against action. Fischer and Ravizza break this sensitivity into two components.

Receptivity to reasons is the ability to recognize reasons as reasons. An agent whose mechanism is receptive to reasons can see that a given consideration counts in favor of or against a particular course of action. She might not always act on those reasons, but she registers them, understands their force, and can weigh them against one another.

Reactivity to reasons is the ability to actually respond to recognized reasons by acting differently. An agent who is reactive to reasons doesn't just see that she has reason to do otherwise; in at least some scenarios, she can actually translate that recognition into a different course of action.

The calibration here is important, and Fischer is careful about it. He argues for moderate reasons-responsiveness, and the reasons for the modifier are philosophically significant.

A requirement of strong reasons-responsiveness would demand that the agent always respond to sufficient reasons. But this is too demanding, because it would strip moral responsibility from anyone who acts against her better judgment. Weakness of will is a real phenomenon. The person who eats the cake knowing full well she shouldn't, the student who procrastinates despite knowing the deadline is tomorrow, the addict who takes the drink while clearly recognizing the reasons not to: all of these agents fail to respond to sufficient reasons, yet we do not for that reason alone deny they are morally responsible. If anything, the weakness of will cases are paradigmatic cases of moral responsibility. We blame the procrastinator precisely because she knew better. Strong reasons-responsiveness sets the bar too high.

A requirement of weak reasons-responsiveness, on the other hand, would demand only that the agent respond to reasons in at least one possible scenario. But this is too permissive. Even severely impaired agents, agents suffering from compulsions, delusions, or deep cognitive dysfunction, might meet this bar, since there might be some far-fetched possible world in which even they respond to a reason. If the bar is that low, then reasons-responsiveness fails to do the work Fischer needs it to do, which is to distinguish responsible agents from non-responsible ones.

Moderate reasons-responsiveness threads the needle. It requires regular receptivity to reasons (the agent consistently recognizes relevant reasons across a range of scenarios) combined with weak reactivity (the agent can act on those reasons in at least some scenarios). This combination captures the intuitive range of responsible agents. The ordinary thief, who sees perfectly well that stealing is wrong but does it anyway, qualifies: he is regularly receptive to moral reasons and could, in some scenarios, refrain. The kleptomaniac, driven by a compulsion she cannot regulate even when she recognizes it, does not qualify: her mechanism is not regularly receptive to reasons in the right way, or if it is receptive, it is never reactive.

We can state this more precisely:

An agent's action issues from a moderately reasons-responsive mechanism if and only if:

(1) The mechanism that actually produces the action is regularly receptive to reasons: across a suitable range of counterfactual scenarios, it recognizes relevant reasons for and against action.

(2) The mechanism is weakly reactive to reasons: there is at least one scenario in which, holding the mechanism's type fixed, the agent acts on a recognized reason and does otherwise.

Mechanism Ownership

Reasons-responsiveness is necessary but not sufficient. Fischer adds a second condition: the mechanism must be the agent's own. This is the condition of mechanism ownership, and it is meant to handle a class of cases where someone is acting from a perfectly reasons-responsive mechanism that was, in some problematic way, installed in them rather than developed by them.

The thought experiment that motivates this condition is the brainwashed agent. Suppose a neuroscientist captures you, rewires your brain, and gives you an entirely new set of values, dispositions, and reasoning patterns. After the procedure, your new mechanism is beautifully reasons-responsive. You recognize moral reasons with sensitivity and nuance. You can act on them. By the lights of moderate reasons-responsiveness alone, you qualify as morally responsible. But something seems wrong. These aren't your values. This isn't your reasoning. It was imposed on you from the outside.

Fischer and Ravizza's solution is to require that the agent take responsibility for the mechanism. This involves a kind of reflective self-identification: the agent sees the mechanism as her own, endorses it (at least implicitly), and has a moderately accurate picture of what the mechanism involves. The idea draws on a broadly Strawsonian framework, following P.F. Strawson's influential 1962 paper "Freedom and Resentment," in which Strawson argued that moral responsibility is grounded in the reactive attitudes (resentment, gratitude, indignation, love) and that these attitudes are appropriate toward agents who are proper participants in interpersonal relationships. Fischer and Ravizza build on this by specifying the conditions under which someone counts as a proper participant: she must act from her own moderately reasons-responsive mechanism.

The Objections

Semi-compatibilism is an elegant and powerful theory. But it faces serious pressure from multiple directions. The outline for the video covers three major objections. I want to expand on those and add a few that the philosophical literature has pressed with particular force.

Objection 1: The Manipulation Problem

This is perhaps the most discussed objection to any actual-sequence account of moral responsibility, and it has been pressed most forcefully by Derk Pereboom in his Living Without Free Will (2001) and later in Free Will, Agency, and Meaning in Life (2014).

Fischer's ownership condition is designed to handle brainwashing cases. But Pereboom constructs a sequence of four cases, each one slightly less "manipulative" than the last, designed to show that Fischer's account cannot draw a principled line between manipulation and ordinary deterministic causation. Here is a simplified version:

Case 1: A neuroscientist directly manipulates Plum's brain in real time, causing him to commit a crime. Plum's mechanism is reasons-responsive and he "owns" it in Fischer's sense.

Case 2: The neuroscientist programs Plum at the start of his life with a set of values and dispositions that will, through ordinary psychological processes, lead to the crime. Same reasons-responsiveness, same ownership.

Case 3: Plum is raised in a community that was intentionally designed by social engineers to produce people with these values and dispositions. Same reasons-responsiveness, same ownership.

Case 4: Plum is an ordinary person in a deterministic universe, raised in an ordinary community, whose values and dispositions are the product of ordinary causal processes stretching back to the Big Bang.

Pereboom's challenge is devastating in its simplicity: if Plum is not responsible in Case 1 (and most people's intuitions say he isn't), and if there is no principled difference between the cases, then Plum is not responsible in Case 4 either. And Case 4 just is what it's like to be a normal person in a deterministic universe.

Fischer's response is to insist that what matters is the actual operation of the mechanism, not its causal history. If the mechanism is reasons-responsive and the agent genuinely owns it, that is sufficient for moral responsibility, regardless of how the mechanism came to be. In his My Way: Essays on Moral Responsibility (2006), Fischer argues that we don't ordinarily trace responsibility back through the full causal history of a person's character. We assess responsibility based on how the character operates now, not on how it was formed.

This is a defensible reply, but it has a cost. It commits Fischer to saying that a perfectly engineered agent, designed from scratch by a neuroscientist down to every value and disposition, is morally responsible for her actions so long as her mechanism is reasons-responsive and she takes ownership of it. Many philosophers find that result deeply counterintuitive. If someone literally made you who you are in every relevant respect, in what sense are you the source of your actions rather than the person who designed you? Alfred Mele, in Free Will and Luck (2006), has pushed a related worry by distinguishing between agents who are "compelled" and agents who are "constituted" by manipulation, arguing that constitution-based manipulation undermines responsibility even when the resulting mechanism looks, from the inside, perfectly healthy.

Objection 2: Can You Really Split Free Will from Moral Responsibility?

The deepest objection to semi-compatibilism is also the most basic. Fischer's entire project depends on the claim that free will and moral responsibility are separable: that you can lack the ability to do otherwise and yet still deserve praise or blame. But is that split really coherent?

The worry runs like this:

(1) Moral responsibility involves genuinely deserving praise or blame.

(2) You can only deserve blame for an action if the action was, in some meaningful sense, up to you.

(3) If determinism is true and you could not have done otherwise, then the action was not up to you in the relevant sense.

(4) Therefore, moral responsibility requires the ability to do otherwise.

(5) Therefore, the split between free will and moral responsibility is incoherent.

Fischer's reply points back to Frankfurt cases. Jones seems responsible for voting for candidate A even though Black's device guaranteed he couldn't have voted for B. If our intuitions confirm Jones's responsibility, then premise (3) is false, and the split is coherent.

But defenders of PAP have a counter-reply, and it deserves attention. The reason we judge Jones responsible, they argue, is that we're tracking the fact that he acted on his own reasons, from his own deliberative process, without any actual intervention. In other words, we are responding to features of the actual sequence. But this doesn't show that alternatives don't matter for responsibility; it might show only that our intuitions about the Frankfurt case are tracking guidance control without actually adjudicating the question of whether alternatives are necessary. We might be confused about whywe think Jones is responsible. The dilemma defense, pressed by David Widerker in "Libertarianism and Frankfurt's Attack on the Principle of Alternative Possibilities" (1995) and developed further by Robert Kane and others, argues that in any coherent version of a Frankfurt case, the agent either retains a morally relevant alternative or the case presupposes determinism in a way that begs the question against the incompatibilist.

This remains one of the most contested points in the literature. Whether Frankfurt cases genuinely refute PAP or merely shift the burden of proof is a question that, after more than fifty years of debate (Frankfurt's original paper, "Alternate Possibilities and Moral Responsibility," appeared in 1969), still does not have a consensus answer.

Objection 3: The Desert Problem

Even if guidance control is sufficient for some form of moral responsibility, is it sufficient for the kind that matters most? Specifically, is it sufficient for desert: the claim that a person genuinely deserves punishment, reward, praise, or blame?

Fischer grounds moral responsibility in the reactive attitudes, following Strawson. If it is appropriate to feel resentment toward someone, it is appropriate to hold them responsible. And guidance control, Fischer argues, makes the reactive attitudes appropriate. If Jones voted for A from his own reasons-responsive mechanism, it makes sense to praise or blame him for his vote, regardless of whether he could have voted differently.

But there is a gap between reactive attitudes and desert, and several philosophers have driven a wedge into it. You can feel resentful toward someone who wrongs you even if you come to believe she couldn't have helped it. The resentment might be natural, psychologically inevitable, even deeply human. But does the person deserve the resentment? Desert seems to require something stronger than the mere appropriateness of a psychological response. It seems to require that the agent could genuinely have chosen differently, that the wrongdoing was, at bottom, up to her, and that she selected it from among live options.

This is the line taken by hard incompatibilists like Pereboom, who argues in Free Will, Agency, and Meaning in Life(2014) that we should abandon backward-looking desert entirely and replace it with a forward-looking framework focused on protection, rehabilitation, and moral formation. On Pereboom's view, semi-compatibilism gives us too little: it gives us a simulacrum of moral responsibility that feels right psychologically but lacks the metaphysical foundation (genuine alternative possibilities) needed to ground genuine desert.

Fischer has responded by distinguishing between different senses of desert. In "Semicompatibilism and Its Rivals" (2012), he argues that the reactive attitudes are not mere psychological reflexes but are constitutively tied to moral responsibility. To say that resentment is appropriate toward someone just is to say that she is morally responsible. There is no further fact of "real desert" that the reactive attitudes might fail to capture. This is, in effect, a Strawsonian deflationism about desert: desert is not some deep metaphysical property that exists independently of our practices of holding one another responsible; it just is the appropriateness of those practices.

Whether this deflationary move satisfies is, I think, one of the deepest questions the debate turns on. If you feel the pull of the thought that a person who was determined to act badly by forces beyond her control doesn't really deserve to suffer for it, no matter how natural our resentment might be, then you'll find Fischer's Strawsonian framework too thin. If, on the other hand, you think that "desert" divorced from our actual practices of praise, blame, gratitude, and resentment is a philosopher's fiction with no real content, then Fischer's view will seem like exactly the right place to land.

Objection 4: The Sourcehood Regress

This objection doesn't appear in the video outline, but it is important enough to include here because it targets the very foundation of Fischer's account: the concept of sourcehood.

Fischer's whole project is built on the idea that what matters for moral responsibility is not whether you had alternatives, but whether you were the source of your action in the right way. Guidance control is supposed to capture the right kind of sourcehood. But some philosophers, notably Galen Strawson (P.F. Strawson's son, interestingly enough), have argued that no coherent notion of sourcehood is available, whether compatibilist or libertarian.

Galen Strawson's "Basic Argument," presented in Freedom and Belief (1986) and refined in numerous subsequent papers, runs as follows:

(1) To be truly morally responsible for what you do, you must be truly responsible for the way you are, since the way you are determines what you do.

(2) To be truly responsible for the way you are, you must have made yourself the way you are, at least in certain crucial mental respects.

(3) To have made yourself the way you are, you must have acted from some prior mental state.

(4) But to be responsible for acting from that prior mental state, you must have been responsible for that prior mental state.

(5) This generates an infinite regress. At some point, you reach a mental state for which you are not responsible, because you didn't create it.

(6) Therefore, true moral responsibility is impossible.

This argument targets everyone, not just Fischer. But it has special force against semi-compatibilism because Fischer explicitly abandons alternative possibilities as a grounding for responsibility and puts all his weight on sourcehood. If the Basic Argument shows that no one is ever truly the source of their own actions, because the source always traces back to something the agent didn't create (genetics, upbringing, the initial conditions of the universe), then Fischer's guidance control is built on sand.

Fischer's reply, consistent with his treatment of the manipulation problem, is to deny premise (1). You do not need to be responsible for who you are in order to be responsible for what you do. Moral responsibility is assessed at the level of particular actions flowing from particular mechanisms, not at the level of ultimate self-creation. The mechanism-ownership condition is Fischer's answer to the worry about self-creation: you don't need to have created your mechanism from scratch; you just need to have endorsed it and taken it as your own.

Whether this is sufficient is, again, a matter of deep philosophical disagreement. Galen Strawson has called this kind of move a "refusal to face the real difficulty," and hard incompatibilists tend to agree. But Fischer and his allies argue that the demand for ultimate self-creation is simply too strong. No one, in any possible world, deterministic or indeterministic, could satisfy it, which means the Basic Argument proves too much: it rules out moral responsibility not just under determinism but under any metaphysical framework whatsoever, including libertarianism. If that's right, then the Basic Argument doesn't tell us anything interesting about determinism versus indeterminism; it tells us that moral responsibility is impossible full stop, which most of us take as a reductio of the argument rather than a reason to accept its conclusion.

Objection 5: The Asymmetry Problem

Here is one more objection worth noting, because it raises a problem internal to Fischer's framework. Fischer has argued, particularly in My Way (2006), for an asymmetry thesis: the conditions for moral blameworthiness and the conditions for moral praiseworthiness might not be the same. Specifically, he suggests that Frankfurt-style cases work more cleanly for blame than for praise. It seems clear that Jones is blameworthy for voting for A if Black's device was standing by to ensure he voted for A anyway. But is Jones praiseworthy for voting for A under those same conditions? Some philosophers have argued that praiseworthiness requires a kind of "going above and beyond," a selection from among genuine alternatives, that blameworthiness does not. If this is right, then semi-compatibilism might deliver moral responsibility for bad actions but not for good ones, which would be a strange and lopsided result.

Fischer has responded by arguing that the asymmetry, to the extent it exists, can be accommodated within his framework. But the point is worth flagging because it suggests that guidance control might not be a fully general account of moral responsibility. It might work beautifully for blame while leaving praise undertheorized.

Where Does This Leave Us?

Semi-compatibilism is one of the most carefully constructed positions in the free will debate. Fischer's move to decouple moral responsibility from the ability to do otherwise is genuinely creative, and the theory of guidance control he builds to replace alternative possibilities is worked out with impressive rigor and sensitivity to counterexamples. It handles weakness-of-will cases elegantly, it draws a plausible line between responsible agents and non-responsible ones, and it avoids the notoriously difficult task of defending alternative possibilities against the Consequence Argument.

But the objections have real weight. The manipulation problem raises the specter that guidance control is too permissive, counting engineered agents as responsible when they shouldn't be. The question of whether free will and moral responsibility can really be split remains genuinely open. The desert problem asks whether guidance-control-based responsibility is thick enough to ground genuine desert or whether it gives us only a sophisticated psychological framework that falls short of what justice requires. And the sourcehood regress suggests that the very concept Fischer builds on may be more fragile than it appears.