Designing Comprehensible Self-Organising Systems

Self-organising systems are a popular engineering concept for designing decentralised autonomic computing systems. They are able to find solutions in complex and versatile problem domains, but as they capture more complexity in their own design, they are becoming less and less comprehensible to their users (be they humans or intelligent agents). We describe a design challenge that relates to usability theory in general and in particular resembles an observation made by Phoebe Senger, who noted that software agents tend to become incomprehensible in their behaviour as they grow more complex. In the manifestation of self-organising systems, the problem is more urgent (since we find ourselves using them more and more) and harder to solve at the same time (since these systems are not centrally controlled). We describe the problem domain and propose three system properties that could be used as quality indicators in this regard: Stability, Learn ability and Engage ability. We demonstrate their usage in a simple model of dynamic pricing markets (e.g. the electricity domain) and evaluate them in different ways.