The Child in the Road: Moral Luck and the Limits of Responsibility

The Hook: Two Drivers, One Coincidence

Picture two people who drink too much on a Friday night and get behind the wheel anyway. Both are equally drunk, both drive equally recklessly, both take the same route home. The first arrives without incident, parks in the garage, goes to bed, and wakes the next morning with nothing worse than a hangover. The second drives down an identical street—except that at that precise moment a child runs after a ball. He cannot brake in time. The child dies.

Now ask the decisive question: Is the second driver a worse person than the first? Does he deserve more moral condemnation?

Almost everyone hesitates here. Two deeply held convictions collide. On the one hand, our gut says unmistakably: yes, the man who killed belongs in court, we despise him, he is guilty. On the other hand, an equally deep principle says: the only difference between the two was the position of a playing child at a fraction of a second—something entirely outside both drivers' control. Both made the same decision, both had the same disposition, both took the same risk. How can pure chance turn one into a killer and the other into a lucky man?

This puzzle goes by the name moral luck. It was thrown into the debate in 1976 by two philosophers, Bernard Williams and Thomas Nagel, in two essays of the same title, and it has haunted moral philosophy ever since. The reason for its explosive force is simple: it shows that one of the most basic assumptions of our morality—that we are responsible only for what is within our grasp—stands in irresolvable contradiction with our actual practice of judgment. And unlike the trolley problem, which constructs a rare extreme situation, moral luck touches every action we ever perform. By the end of this article it will become clear that this abstract debate leads straight into the engine room of our professional lives: into the question of how we judge engineers, security leads, and ourselves after failures whose outcome we could never fully control.

The Core Concept: The Control Principle

To understand why moral luck is a problem and not merely a curiosity, you have to know the conviction it violates. Nagel states it with great clarity. He calls it an intuition we share "prior to reflection":

"Prior to reflection it is intuitively plausible that people cannot be morally assessed for what is not their fault, or for what is due to factors beyond their control." (Nagel, 1979)

This idea is known in the literature as the control principle: We are morally assessable only to the extent that what we are assessed for depends on factors under our control. From it follows a natural corollary: if two people are alike in everything that lies within their control, then they cannot differ in anything morally relevant.

The control principle is no philosophical exotic. It is the beating heart of justice as we feel it. It lies behind the outrage that someone should be penalized for their skin color, their origin, or their sex—all things no one chooses. It is why we distinguish murder from manslaughter, intent from accident. And in its purest form it appears in Immanuel Kant. For Kant, the only unconditionally good thing in the world is the good will, and the good will, as he writes, "shines like a jewel by itself"—entirely independent of whether it achieves its aim. Whether your good intention succeeds or is thwarted by adverse circumstances makes no difference to its moral worth. Morality, on the Kantian ideal, is the one sphere in which fate holds no power.

It is precisely this consoling picture—Williams calls it "solace to a sense of the world's unfairness"—that the two philosophers target. Their thesis: the control principle is deeply anchored in us, but we cannot sustain it. As soon as we look honestly, we find that we constantly praise and blame people for things that depend on luck and chance. And if we try to exclude every luck-touched action from our moral judgment, what remains in the end is—almost nothing.

Part 1: Williams and the Case of Gauguin

Bernard Williams approached the problem from a surprising angle. His real aim was to shake the Kantian conception of morality altogether and replace it with a wider notion he called "the ethical." His lever was the concept of justification.

Williams tells a story loosely modeled on the life of the painter Paul Gauguin. His fictional Gauguin lives reasonably contentedly with his wife and children, feels a genuine responsibility toward them—and abandons them nonetheless. He sets off for the South Seas, convinced that only there, far from civilization, can he ripen into a truly great painter. In doing so he plunges his family into misery.

Was that decision justified? Williams' crucial move is the observation that this simply cannot be answered at the moment of decision. Whether Gauguin was right depends on whether he actually becomes a great artist—and no one, not even Gauguin himself, can foresee that. In Williams' famous phrase: "the only thing that will justify his choice will be success itself." If the work succeeds, Gauguin's break with his family is justified in retrospect. If he fails, it was the blunder of an egotist who sacrificed his family for an illusion. Justification is determined retroactively by an outcome the agent did not hold in his hands.

Williams refines this further with an important distinction. Not every success justifies, not every failure condemns. It depends on the kind of luck. He separates intrinsic from extrinsic luck. Intrinsic luck arises from the project itself—for instance, whether Gauguin possesses the talent and inner strength he credits himself with. Extrinsic luck comes from outside: if Gauguin breaks his wrist on the way to the ship and can never paint again, then his project has failed, but no verdict has been passed on his original decision—the accident had nothing to do with the matter. Only intrinsic luck, Williams argues, can determine whether a choice was reasonable in hindsight.

The real assault on Kant comes in the next step. Suppose Gauguin's decision to leave his family is morally unjustified. On the Kantian understanding, that is fixed at the moment of decision and stays that way forever—luck has no influence on moral worth. Rational justification, however, as Williams has shown, hangs on the outcome. So two kinds of justification come apart: Gauguin's act can be morally reprehensible and at the same time rationally justified. And if morality does not automatically win such a conflict—if we end up admiring Gauguin for the courage to stake everything on one card—then morality is not the highest, all-surpassing source of value Kant took it to be. Williams' dilemma runs: either moral worth is sometimes a matter of luck after all, or it is not the supreme value. Either horn destroys the Kantian picture.

Williams' argument is still regarded as the weaker of the two—not least because it hangs on the idiosyncratic and contested notion of rational justification through success. But it put a finger on the wound. The clean, systematic incision was then made by Thomas Nagel.

Part 2: Nagel's Four Types of Luck

Nagel defines the phenomenon precisely:

"Where a significant aspect of what someone does depends on factors beyond his control, yet we continue to treat him in that respect as an object of moral judgment, it can be called moral luck." (Nagel, 1979)

His lasting contribution is a map. Nagel shows that luck creeps not only into the results of our deeds but into every layer of agency—down to the very person we happen to be. He distinguishes four types (the catchy names come partly from later authors such as Daniel Statman and Michael Zimmerman):

1. Resultant Luck

"Luck in the way one's actions and projects turn out." This is the case of the two drunk drivers, and also of Gauguin's success. Two people act identically, but the world decides differently about the outcome—and we judge accordingly. Nagel's second example is political: whoever instigates a revolution against a brutal regime becomes a celebrated hero if it succeeds and bears "some responsibility" for the bloodbath if it fails—even though success depends on a thousand uncontrollable factors.

2. Circumstantial Luck

The luck of "the kind of problems and situations one faces." Nagel's most piercing example: a citizen in 1930s Germany faces moral tests he may fail or pass heroically. A German who emigrated to Argentina in 1930 and led a quiet business life there was never confronted with those tests. Perhaps he would have behaved just as opportunistically—or just as courageously—but he was never tried. His moral "record" stays clean, solely because fate removed him from the danger zone. We are all morally as good or bad as the situations we happen to fall into allow.

3. Constitutive Luck

The luck of having "the inclinations, capacities, and temperament" one has. This is the deepest and most unsettling type. Whether you have a hot-tempered or a gentle nature, whether courage, generosity, and self-control come easily to you or whether you must fight against envy, cowardice, and rage—all of that is largely the product of your genes, your childhood, your formation. You chose none of it. And yet we blame the cruel man for his cruelty and praise the kind one for his kindness, as though they had given themselves their own characters.

4. Causal Luck

"Luck in how one is determined by antecedent circumstances." This is at bottom the ancient problem of free will in new dress: if everything we do is the result of an unbroken causal chain that began before our birth, then ultimately none of our actions is truly "ours." Nagel himself suggests that this fourth type merges with the other three—since circumstance and constitution are just special cases of being "determined by antecedent circumstances."

The combined effect of these four types is devastating. Nagel states the point relentlessly:

"Ultimately, nothing or almost nothing about what a person does seems to be under his control." (Nagel, 1979)

Result, situation, character, indeed the causal origin of every impulse—layer after layer is stripped away, and if we apply the control principle consistently, no acting "self" remains at the end to which anything could be ascribed at all. The moral subject "shrinks to an extensionless point," as Nagel puts it.

Part 3: The Paradox—Why This Is No Mere Puzzle

Nagel draws an illuminating parallel to epistemological skepticism. The skeptic argues: real knowledge would have to be certain and non-accidental—but hardly any of our true beliefs are entirely free of accident, so strictly speaking we know almost nothing. Precisely the same erosion strikes morality. Just as with the Gettier problem, where epistemic luck prevents a true justified belief from being genuine knowledge, here moral luck undermines the clean ascription of responsibility. In both cases the lesson is the same: a principle reasonable in itself, thought through to the end, devours itself.

That is Nagel's real diagnosis—and the reason moral luck is not a mere riddle to be dissolved with a clever trick. It is a paradox at the heart of morality itself. We can neither abandon the control principle (then we would have to blame people for birthplace and the genetic lottery, which would be deeply unjust) nor sustain it consistently (then no object of judgment would remain). Both ends are uninhabitable, and yet we stand with one foot on each. I am of the opinion that precisely this uninhabitability is why the debate is not closed after fifty years—not because philosophers are too slow, but because a genuine fault line runs through our concepts here.

Part 4: The Experimental-Philosophy Turn

For a long time moral luck was pure armchair philosophy: you consulted your own intuitions. But since roughly 2010, experimental philosophy has brought the problem into the laboratory and had thousands of subjects actually render judgments. The results complicate the picture in fascinating ways.

First, the outcome effect (outcome bias) was massively confirmed: people judge an agent more harshly when his deed turns out badly, even when intent and risk were identical. The manslaughter driver is reliably condemned more severely than his lucky twin. So far, so expected.

But then came the fine resolution. A widely noted study by Markus Kneer and Edouard Machery ("No Luck for Moral Luck," published in 2019 in the journal Cognition) revealed a decisive difference: if you ask people in a between-subjects design—each subject sees only one of the two cases—they judge the unlucky driver more harshly. But if you confront the same people in a within-subjects design with both cases side by side, so they can make the comparison, then the difference in judgments of blame, wrongness, and permissibility of the action largely disappears. The overwhelming majority then hold the lucky and unlucky drivers equally blameworthy. In other words: as soon as people reflect carefully and comparatively, they commit to the control principle. The outcome effect thus looks less like a stable moral conviction and more like a judgment error we fall for when we have only one case in isolation before us.

Especially telling is which judgments hang most strongly on the outcome. Several studies (among them a large 2023 investigation, "Outcome effects, moral luck and the hindsight bias," in Cognition) agree: punishment judgments are far more outcome-dependent than judgments of wrongness or blame. We punish the actual harm, even when we rate the disposition the same. And the mechanism behind it is the hindsight bias: once the harm has occurred, it seems to us in retrospect more probable than it was beforehand. As a result we attribute more negligence to the agent—"he should have seen it coming!"—and through this inflated negligence the outcome seeps into our verdict of blame. Luck thus does not directly alter our morality; it alters our perception of the risks, and the distorted risk perception then alters the judgment.

This is a remarkable finding. It suggests that a good part of apparent moral luck is in truth a cognitive error—not a deep feature of morality but a predictable distortion of our brains, akin to the statistical artifacts that also partly exposed the Dunning-Kruger effect as a measurement error. Yet the paradox is not thereby fully dissolved: constitutive and causal luck cannot be so easily explained away as perceptual mistakes.

Part 5: The Attempts at Resolution

So how to handle the paradox? The answers of recent decades sort into two broad camps.

Camp 1: The intuition is mistaken—luck may make a moral difference. This is the smaller camp. It argues that the control principle is a leftover of an all-too-Kantian, "pure" conception of moral agency. Aristotelians, for instance (whom the control principle scarcely troubles), see morality as a matter of the flourishing life and of character proving itself in a world full of chance—vulnerability to luck belongs, for them, to human virtue itself. Martha Nussbaum, in The Fragility of Goodness (1986), showed how deeply Greek tragedy already circles precisely this vulnerability of the good. For those who think this way, moral luck is not an anomaly but a basic fact of ethical life.

Camp 2: The "fact" is not one—luck never really makes a moral difference. This is the more popular strategy. It claims: the lucky and the unlucky driver are morally exactly equal; the difference is purely epistemic. The death that occurred did not change the unlucky driver's character—it merely made it visible. The lucky driver is just as guilty; he simply has the good fortune that his guilt goes undetected (even by himself). On this reading, if we knew everything, both drivers would deserve equally harsh condemnation. Michael Zimmerman has spelled out this position sharply with his notion of "luck-free moral assessment." Notably, the experimental findings from Part 4 give this camp a tailwind: if people, reflecting comparatively, really do return to the "equally guilty" verdict, then resultant luck may indeed be only an epistemic illusion.

But this second camp has an open flank, and it lies precisely at constitutive luck. The epistemic rescue works well for resultant and circumstantial luck: there, chance merely conceals or reveals an already existing character. But what if the character itself is a product of luck? If someone possesses, through the sheer lottery of genes and upbringing, a cruel, deceitful character—what pre-existing, luck-free feature does chance reveal to us then? Here the epistemic strategy grasps at nothing. Some philosophers draw the radical conclusion that the notion of constitutive luck is itself incoherent; others see in it the proof that the paradox is, after all, irresolvable. I am of the opinion that this is the most honest point in the whole debate: resultant luck can perhaps be explained away, constitutive luck cannot—and so a hard, irreducible core of the problem remains.

Part 6: From the Seminar Room to the Engine Room

Why should an engineer, a cloud architect, or a security lead care about any of this? Because moral luck does not live in South Sea studios and trolley tracks but in every deployment, every code review, every risk decision.

Consider two developers. Both write the same negligent code—say, an unvalidated input that permits SQL injection. For one, an attentive colleague catches the bug in review, it is quietly fixed, no one ever speaks of it. For the other, the same bug slips into production, an attacker finds it, a data breach follows, and the company is in the headlines. The second developer may lose her job; the first gets a promotion. The only difference was the timing of another person's glance and the attention of an attacker—pure resultant luck. Anyone who knows the story of the XZ backdoor knows how razor-thin the line between "near-global catastrophe" and "caught in time" often is—there, the accidental curiosity of a single developer, over half a fraction of a second of latency, made the difference.

It is precisely here that the IT industry, often without knowing the philosophical name, has developed one of the most mature practical answers to the problem of moral luck: the blameless post-mortem culture. Its core idea is essentially pure control principle: assess the decision in the light of what the agent could know and control at the time of the decision—not in the harsh hindsight glare of the actual outcome. Blameless culture is thus a direct antidote to the hindsight bias of Part 4: it forces us to dismantle the inflated, retrospective attribution of negligence and to ask whether the mistake was recognizable as negligent beforehand as well.

From this a compact assessment grid for everyday work can be drawn:

And the debate reaches to the current frontline: who bears responsibility when an AI system causes harm? The developer could never fully control the specific output; the model is itself a product of constitutive luck from training data and random initialization. The same questions Nagel put to the drunk driver return here in sharpened form—and the ethics of autonomous systems, which we already touched on with the trolley problem, will not manage without a clear concept of moral luck.

The Central Takeaway

The most important practical lesson of moral luck is an exercise in double-entry bookkeeping of judgment. Deliberately separate two questions our brain constantly conflates: How bad is the outcome? and How blameworthy was the decision? The outcome may and must guide our action—we repair the damage, we learn, we improve. But it should not retroactively inflate our moral verdict on the person whose disposition and knowledge were not changed one iota by the result.

Concretely: the next time you condemn someone (or yourself) for a mishap, imagine their lucky twin—someone who did exactly the same thing but to whom chance was merciful. Does this twin really deserve less blame? If not, then your outrage is aimed not at a decision but at the throw of the dice. This small mental swap is the sharpest tool against hindsight bias—and at the same time a source of mercy, toward others and toward oneself.

An Open Question to Ponder

Suppose you accept that no one chose their own character—that courage and cowardice, generosity and hardness are largely gifts or burdens of constitutive luck. Does any meaningful place then remain for genuine moral desert and genuine guilt—or is our entire web of praise and blame, pride and shame, in the end merely a useful fiction we maintain because a society without it would not function?

Cross-References in the Vault

• The Switch in the Machine: The Trolley Problem and the Ethics of Autonomous Vehicles – the other great construction site of applied ethics: when machines decide over life and death, moral luck becomes a question of design.
• Three Pages Against 2,000 Years: The Gettier Problem and What Knowledge Really Is – the epistemological twin problem: there epistemic luck undermines knowledge, here moral luck undermines responsibility.
• What Mary Did Not Know: The Knowledge Argument and the Riddle of Consciousness – likewise a thought experiment that shakes a deeply held principle (physicalism) with a single case.
• The Two Boxes: Newcomb's Paradox and the Battle Over Rational Choice – another paradox of rational choice, where predictability and control come into tension.
• The Peak of Not Knowing: Why Psychology's Most Famous Effect May Be Just a Statistical Mirage – how a seemingly deep psychological truth turns out to be partly an artifact of measurement; kin to the hindsight bias in moral luck.

Sources

• Internet Encyclopedia of Philosophy: Moral Luck – https://iep.utm.edu/moralluc/
• Stanford Encyclopedia of Philosophy: Moral Luck – https://plato.stanford.edu/entries/moral-luck/
• Thomas Nagel, "Moral Luck," in: Mortal Questions, Cambridge University Press, 1979 (originally Aristotelian Society Supplementary Volume, 1976).
• Bernard Williams, "Moral Luck," in: Moral Luck: Philosophical Papers 1973–1980, Cambridge University Press, 1981 (originally 1976).
• Kneer, M. & Machery, E. (2019): "No Luck for Moral Luck," Cognition 182, pp. 331–348 – https://www.sciencedirect.com/science/article/abs/pii/S0010027718302403
• "Outcome effects, moral luck and the hindsight bias" (2023), Cognition – https://www.sciencedirect.com/science/article/pii/S0010027722002463
• "Moral luck and the roles of outcome and negligence in moral judgments" (2023), Journal of Experimental Social Psychology – https://www.sciencedirect.com/science/article/pii/S0022103123000136
• Wikipedia: Moral luck – https://en.wikipedia.org/wiki/Moral_luck

← All articles