CONCEPT
Moral Luck (Williams)
Williams's 1976 thesis that moral assessment depends on factors beyond the agent's control — an uncomfortable fact that threatens every system presupposing a tight connection between desert and choice, and that the AI transition has made unignorable.
Moral luck names the phenomenon that identical actions, performed by agents of identical moral quality, can receive radically different moral assessments because of outcomes neither agent controlled. Williams's paradigmatic case is two drivers who run a red light: one strikes a pedestrian, one arrives safely. The moral-philosophical tradition, particularly its Kantian strand, insists that moral worth attaches only to the will, which is insulated from contingency. Williams regarded this position as elegant and false. We do not live in the kingdom of ends. Outcomes matter, identities are partly constituted by what happens to us, and the AI transition has generated a vivid new species of moral luck by distributing gains and losses across practitioners according to which skills happened to align with the technology's capabilities.
In The You On AI Field Guide
Williams identified four varieties of moral luck (elaborated by Nagel in the companion 1976 paper): constitutive luck in the kind of person one is,