If our aim is to build morality into an artificial agent, how might we begin to go about doing so?