Assignment of Legal Responsibilities for Decisions by Autonomous Cars Using System Architectures

Abstract

Increasing automation in road vehicles moves decision making from humans to algorithms. This gives a lack of clarity in assigning responsibility for the consequences of the resultant actions, presenting problems for ethicists, lawyers, and engineers. This article applies proven methods based on the 4D/RCS reference model architecture to derive proposed architectures for autonomous vehicles with J3016 autonomy levels 2-5. The architectures have a hierarchy of nodes, which can be either a human or an automated process, giving unambiguous allocation of authority to act and consequent responsibilities to individual nodes in the system. A concept called authorized power is introduced for every node, giving clear limits on each node's ability to act. Setting node requirements gives a firm basis for engineers and lawyers to agree where responsibilities lie, whether with the user, owner, or in the supply chain. It also provides a framework to assess engineering risk and the wider financial implications of the new technologies as they are introduced.