Towards an Ethical Machine: One test at a time

Abstract Machines acting as artificial agents are increasingly operating in environments populated by human beings, and are interacting directly with people. In order for artificial agents to be accepted by society, they must act in a manner that reflects society's norms. Since a society's norms are grounded in ethical rules of behaviour, artificial agents will also need to follow ethical norms of the society in which they operate, if they are to be accepted by human agents. In addition to an artificial agent's goals, tasks and actions for which it is designed such as operating as a care giver or an autonomous vehicle, an artificial agent needs to also incorporate ethical considerations when choosing which task to perform. To achieve this, an ethical component needs to be added to the artificial agent's programming. I present one such approach to achieve this based on Kent Beck's Test-Driven Development methodology