Moral luck describes circumstances whereby a moral agent is assigned moral blame or praise for an action or its consequences, even if it is clear that said agent did not have full control over either the action or its consequences. This term, introduced by Bernard Williams, has been developed, along with its significance to a coherent moral theory, by Williams and Thomas Nagel in their respective essays on the subject.

Broadly speaking, human beings tend to correlate, at least intuitively, responsibility and voluntary action. Thus, the most blame is assigned to persons for their actions and the consequences they entail when we have good cause to believe that both:

Conversely, there is a tendency to be much more sympathetic to those who satisfy any of the following conditions:

Parenthetically, the above criteria do not correlate exactly with moral praise – while it may be true that one can, and should assign a good deal of moral praise to those who had performed a good action, or an action entailing good consequences, completely on their own volition and uncoerced, it is debatable that the same distinction holds for involuntary actions that happened to turn out well or happened to produce good outcomes.

This correlation between responsibility and voluntary action is acceptable to most people on an intuitive level; indeed, this correlation is echoed in American and European law: for this reason, for example, manslaughter, or killing in self-defense carries a significantly different type of legal punishment (i.e., formalized moral blame) than premeditated murder.

Given the notion of equating moral responsibility with voluntary action, however, moral luck leads to counterintuitive solutions. This is illustrated by an example of a traffic accident. Driver A, in a moment of inattention, runs a red light as a child is crossing the street. Driver A tries to avoid hitting the child but fails and the child dies. Driver B also runs a red light, but no one is crossing and only gets a traffic ticket.

If a bystander is asked to morally evaluate Drivers A and B, they may assign Driver A more moral blame than Driver B because Driver A's course of action resulted in a death. However, there are no differences in the controllable actions performed by Drivers A and B. The only disparity is an external uncontrollable event. If it is given that moral responsibility should only be relevant when the agent voluntarily performed or failed to perform some action, Drivers A and B should be blamed equally. This may be intuitively problematic, as one situation resulted in a death.

Nagel (1979) identified four kinds of moral luck in his essay. The kind most relevant to the above example is "resultant moral luck".