"Know thyselves" - Computational Self-Reflection in Collective Technical Systems

The domain of self-adaptive and self-organising systems has faced noticeable attention within the last decade, since it investigates solutions to tackle complexity challenges arising from the increasingly coupled and dynamic character of emerging technical systems. The resulting solutions react to changing conditions and self-optimise their decisions over time. In this article, we outline that future intelligent technical systems will have to act as a collective, and this collective has to be equipped with novel techniques for decision strategies. Thereby, we have to go far beyond the existing reactive approaches in terms of proactive modelling of knowledge and goals, a continuous evaluation of goal achievement, and a dynamic goal adaptation process - to which we will refer to as "collective selfreflection". We provide a definition of this term, an architectural blueprint, and a draft of a research agenda towards collective self-reflection. We introduce an application scenario called "swarm fleet infrastructure" to motivate the need of such techniques.