A Cognitive Architecture for Verifiable System Ethics via Explainable Autonomy

摘要

Immersing autonomous technologies into our technical and social infrastructures has profound implications and hence requires instilling confidence in their behavior to avoid potential harm. In this paper, we present Cogent, an interactive coherence-driven agent specification and simulation platform and demonstrate its use in ethical decision-making. To advance the theory, methodology, and applications of autonomous system verification, Cogent leverages a cognitive model to bear on the challenges of reasoning about trust, ethics, and intentions. The underlying strategy of the cognitive model is based on the concept of a connectionist, interactive activation model that implements the theory of coherence. To illustrate the practical utility of Cogent, we explore a machine ethics case study.