Concern processing in autonomous agents.
SR Allen · 2001
Concerns are broadly defined as dispositions to desire the occurrence or non-occurrence of a given kind of situation. In this thesis we present an information-level analysis of the mechanisms that render the concerns of intelligent agency in the symbolic, situated, and “emotional” programming paradigms – to give an account of the functions, constraints and types of concern processes, and to investigate cognitive architectures that are capable of supporting such processes. Part I introduces the research topic and describes the motivated agent framework used within the Cognition and Affect Project, and this thesis, to elucidate the architectural requirements for intelligent autonomous agency. Part II focuses on the issue of concernprocessing in autonomous agency. We identify weaknesses in current deliberative and behaviour-based design approaches, and provide two case studies of our concern-centric information-level design-based approach to intelligent autonomous agent design. Part III applies our design methodology to the requirements for human emotional states. We present a information-level analysis of leading theories of emotion, and describe a series of broad agent architectures for elucidating emergent infant-like emotional states. Part IV summarises the different strands of research presented within this thesis, and identifies some fertile areas for future research. By describing a variety of functions using the design stance at the information-level, and showing how they account for human-like mental states and processes, we aim to provide a rich explanatory framework for intelligent autonomous agency.