Continuation under Incompleteness. Proven Reliability, Settling-In and Branchable Development of Autonomous Systems

This independently readable companion paper extends the conceptual synthesis and discussion paper “Robotics Capable of Development” by examining the temporal development of reliable competences, local settling-in, responsibility and the continuation of autonomous systems under incomplete knowledge. A robot operating for extended periods in an open human environment cannot know every future condition in advance or possess every required competence as a fully specified rule from the outset. The paper therefore asks how abilities are acquired, how their reliability is demonstrated in practice and how the conditions and limits of their validity can remain recognizable during operation. Central distinctions concern competence, proven reliability and authorization, as well as general training and local settling-in. Other topics include historically established reliability, trust, responsible action under uncertainty, the interests of affected third parties, and the possibility of revising established routines when the world changes. A further focus is the continuation of artificial development after interruptions. The paper discusses the preservation of Gewordenheit, selective memory, reconnection after disruptions, backups, changes of physical carrier and branchable artificial ontogeny, in which shared developmental states can give rise to different subsequent learning histories. The paper presents neither a finished technical architecture nor a systematic literature review, and it makes no claim of scientific priority. Its purpose is to connect conceptual distinctions, research questions, possible experiments and counter-hypotheses that may help investigate the conditions of responsible, reliable and developmentally open autonomy.