Moral machines : perception of moral judgment made by machines

Abstract

While technological development of vehicular autonomy has been progressing rapidly, a parallel discussion has emerged with regard to the moral implications of a future wherein people hand over to autonomous machines the controls to a mode of transportation. These discussions have entered a new phase with the U.S. Department of Transportation (DoT) releasing a 15-point policy that requires manufacturers to explain how their AVs will handle "ethical considerations". However, there is a huge gap in our understanding of the ethical perception of Al, as there have been few large-scale empirical studies on human moral perception of outcomes to autonomous vehicle moral dilemmas. Additionally, public engagement is a very important piece of the puzzle, especially given the emotional salience of traffic accidents. With that in mind, I co-developed the "Moral Machine" (http://moralmachine.mit.edu). Moral Machine is a platform for gathering a human perspective on moral decisions made by machine intelligence, such as AVs. The web site went viral, and got covered in various media outlets. This web site has also been a valuable data collection tool, allowing us to collect the largest dataset on Al ethics ever collected in history (with 30 million decisions by over 3 million visitors, so far). This thesis will introduce the Moral Machine platform as a data-gathering platform. Moreover, insights about the human perception of the different routes to full automation will be covered in the thesis, with the data collected through other online platforms.