The Moral Machine Experiment was a large-scale experiment in which participants across the world were asked to make (fictive) decisions in moral dilemmas. Specifically, they were asked about their decisions for various configurations of the trolley problem. The trolley problem involves a vehicle that if it stays on course will kill a group of individuals A currently crossing the road (at a green or red stoplight), or if it swerves will kill a group of individuals B that in one variation are its passengers or in another variation other individuals on the street. The participants of the experiment then decide whether the vehicle should kill the individuals A or B. In the experiment, all groups of individuals were composed of different types and numbers such as "3 female athletes and a cat" vs. "one male doctor". The moral preferences of participants varied greatly and might have been influenced by various factors such as their cultural background.
The student's task is to study the related work that has been done on this data set, and for example:
- find suitable preference models
- employ (and potentially modify) algorithmic solutions to learn preference models from the data
- evaluate the quality of the preference models w.r.t. suitable measures
or
- consider suitable aggregation functions for incomplete preferences
- compute (approximate) aggregations of the participants preferences into a single preference model / order representing the society as a whole
- evaluate the properties of the aggregated preferences.