A Symbiotic Cognitive Computing Perspective on Autonomic Computing

Abstract

Symbiotic Cognitive Systems (SCS) are multi-agent systems comprising both human and software agents that are designed to collectively perform cognitive tasks such as decision-making better than humans or software agents can unaided. Autonomic Computing Systems (ACS) are multi-agent systems that manage applications as well as software and hardware resources in accordance with goals specified by human administrators and users. SCS and ACS share some key characteristics. First, both are designed to extend human intellectual capabilities, and as such they require effective means by which humans can communicate their objectives to the computing system. Second, their natural architecture is a multi-agent system in which dozens, hundreds or even more semi-autonomous entities interact. In both SCS and ACS, issues of inter-agent communication and coordination come to the fore. We report our experience with a moderate-scale SCS prototype that helps human experts make decisions with financial impacts ranging from millions to even billions of US: corporate mergers and acquisitions. Taking advantage of the commonalities, we translate this experience into insights that may benefit future research on ACS, and recommend a stronger focus on agent-human communication and building realistic system prototypes.

DOI: 10.1109/ICAC.2015.16

3 Figures and Tables

Cite this paper

@article{Kephart2015ASC, title={A Symbiotic Cognitive Computing Perspective on Autonomic Computing}, author={Jeffrey O. Kephart and Jonathan Lenchner}, journal={2015 IEEE International Conference on Autonomic Computing}, year={2015}, pages={109-114} }