No abstract

In the moral machine project, participants are asked to form judgments about the well-known trolley example. The project is intended to serve as a starting point for public discussion that would eventually lead to a solution to the social dilemma of autonomous vehicles. The dilemma is that autonomous vehicles should be programed to maximize the number of lives saved in trolley-style dilemmas. But consumers will only purchase autonomous vehicles that are programed to favor passenger safety in such dilemmas. We argue that the project is seriously misguided. There are relevant variants of trolley to which the project's participants are not exposed. These variants make clear that the morally correct way to program autonomous vehicles is not at odds with what consumers will purchase. The project is hugely popular and dominates public discussion of this issue. We show that, ironically, the project itself is largely responsible for the dilemma. Keywords Moral machine project • Trolley problem • Autonomous vehicles In the moral machine project 1 , participants are asked to form judgments about variations of this well-known example: Trolley: There is a runaway trolley. If you do nothing, the trolley will hit and kill five people. If you pull a lever, the trolley will be diverted onto another track and kill one person.

Hume seems to claim that there does not exist a valid argument that has all non-ethical sentences as premises and an ethical sentence as its conclusion. Starting with Prior, a number of counterexamples to this claim have been proposed. Unfortunately, all of these proposals are controversial. Even the most plausible have a premise that seems like it might be an ethical sentence or a conclusion that seems like it might be nonethical. Since it is difficult to tell whether any of these counterexamples are genuine, we need a taxonomy that sorts out ethical sentences from non-ethical ones. We need to know the difference between an 'Is' and an 'Ought'. In the first part of the paper, I establish the need for a taxonomy. I consider some of the most influential 'Is'-'Ought' derivations. These include proposals by Prior and Searle. I argue that each proposal has a premise whose status as ethical or non-ethical is difficult to determine. In the second part of the paper, I consider taxonomies proposed by Karmo and Maitzen. I argue against both taxonomies. I end with the claim that we need a taxonomy of ethical sentences and that none of the current proposals are adequate.

I defend a new version of the non-identity theodicy. After presenting the theodicy, I reply to a series of objections. I then argue that my approach improves upon similar approaches in the literature.