Gödelian Self-Reference in Agent-Oriented Software

The paper aims at substantiating the first steps of a more general undertaking regarding self-awareness in agent-based systems, founded on Hofstadter’s ideas and presented in previous papers (illustrating the broad-band technology potential from an anthropocentric and transdisciplinary perspective). This second paper focuses on computer science aspects, keeping a definite engineering perspective: the target is a generic architecture – based on Gödelian self-reference – for agent-based applications meant for open, heterogeneous, dynamic and uncertain environments. Specific objectives are: a) to defend the undertaking from an agentoriented software engineering point of view; b) to detail the rationale for starting by Gödelian self-reference; c) to specify a stepwise approach affordable within a narrow scope; d) to present self-cloning as the basic software mechanism able to uphold this approach; e) to outline very roughly an experimental model. (Details and implementation issues regarding mechanisms and models are described in forthcoming papers.) Preliminary conclusion: the approach seems workable and the applicative potential seems significant.