Extending Artificial Agency to Incorporate Ethics-oriented Social Interaction

AI is still far away from the implementation of a truly autonomous agent, and so far, attempts to build Autonomous Moral Agent (AMAs) have been concentrated on the ad hoc integration of a moral theory to the mechanisms of an Artificial Agent (AA). However, it would be more natural if morals were to evolve within the cognitive realm of the artificial system rather than function separately. A systemic framework of second order cybernetics and self-organization is proposed as the basis for the design and implementation of artificial AMAs. In this proposal information is not taken in the ordinary sense and ethics are reflected in the quality of information processing by the system. This processing is made through the creation of new meaning structures, based on the intentions and goals of the agent. In this perspective, being an autonomous agent is equivalent to being an AMA. The limitations of the framework from an implementation point of view are indicated and an extension of it via the incorporation of semiotic processes is suggested.