Trustworthy “blackbox” Self-Adaptive Systems

Paper i proceeding, 2023

For humans to trust Self-Adaptive Systems in critical situations, they must be robust, ethical, and lawful, but human intelligence is still needed to make ethical decisions. This paper presents a framework to discuss human values in the RE process for Self-Adaptive Systems and RE-specific challenges arising due to the AI paradigm shift towards foundation models: self-supervised blackboxes. Semi-autonomous heavy mining vehicles are a running example to present the requirements.