Jantsch, A. (2019). Towards a Formal Model of Recursive Self-Reflection. In S. Saidi, R. Ernst, & D. Ziegenbein (Eds.), Workshop on Autonomous Systems Design (ASD 2019)} (pp. 1–15). Schloss Dagstuhl--Leibniz-Zentrum fuer Informatik. https://doi.org/10.4230/OASIcs.ASD.2019.6
Self-awareness holds the promise of better decision making based on a comprehensive assessment of asystem´s own situation. Therefore it has been studied for more than ten years in a range of settingsand applications. However, in the literature the term has been used in a variety of meanings andtoday there is no consensus on what features and properties it should include. In fact, researchersdisagree on the relative benefits of a self-aware system compared to one that is very similar butlacks self-awareness.We sketch a formal model, and thus a formal definition, of self-awareness. The model is basedon dynamic dataflow semantics and includes self-assessment, a simulation and an abstraction asfacilitating techniques, which are modeled by spawning new dataflow actors in the system. Mostimportantly, it has a method to focus on any of its parts to make it a subject of analysis by applyingabstraction, self-assessment and simulation. In particular, it can apply this process to itself, whichwe call recursive self-reflection. There is no arbitrary limit to this self-scrutiny except resourceconstraints
