Overview
Vehicles become more and more complex with the advances in the automated vehicle technology. The deployment of automated vehicle (AV) will take place step by step as the level of vehicle’s autonomy and the complexity of in its environment increases, moving from highway to urban or rural areas. During this transition period, the AV will become a social agent taking decisions to regulate its interactions with other vehicles, vulnerable road users, and static objects in its environment in a mixed traffic. An AV will find itself in unavoidable collision conditions that oblige the vehicle to make real time decisions of risk distribution in ethical dilemma involving high uncertainty. The ethical principles of an AV thus become indispensable to take into account from the early stages of design.
The global objective of the AVEthics project is to produce recommendations for a societally acceptable ethics policy for an AV in ethical dilemma situations. We adopt a multidisciplinary approach to realize our objective, bringing together experts from philosophy, artificial intelligence and robotics, and social psychology. The first aim is to define and specify the kind of (artificial) ethics that can be applied to the artificial intelligence of an AV, including its moral principles, values and weighing rules and their position with respect to human ethics and ontology. The second aim is to model the properties of an artificial ethics numerically in order to test the interactions of an AV with other road users in dilemma situations. The third aim is to evaluate the social acceptability of the ethics principals proposed for an AV applied to simulated use cases.
In line with these goals, the expected outcomes of the AVEthics project are 1) an initial proposition of ethical principles for an AV, 2) a numerical modelling interface to apply these ethical principles to different use cases, and 3) end user’s acceptability judgments of the proposed ethical principles and the following action plans. The outcomes of the project are expected to improve the operational safety design of an AV and render it acceptable for the end-user and the stakeholders of the mobility system.