Emotionally-Informed Decisions: Bringing Gut's Feelings into Self-adaptive and Co-adaptive Software Systems
This addresses the need for more human-like adaptability in autonomous software systems, though it appears conceptual/incremental.
The paper tackles the problem that self-adaptive and co-adaptive software systems lack human-like emotional decision-making capabilities, proposing algorithms for realistic decision-making and architectures that incorporate human emotions to quantify uncertainty and aid adaptation.
Software systems now complement an incredibly vast number of human activities, and much effort has been deployed to make them quasi-autonomous with the build-up of increasingly performant self-adaptive capabilities, so that the burden of failure, interruption and functional loss requiring expert intervention is fewer and far in between. Even as software systems are rapidly gaining skills that beat humans', humans retain greatly superior adaptability, especially in the context of emotionally-informed decisions and decisions under uncertainty; that is to say, self-adaptive and co-adaptive software systems have yet to acquire a "gut's feeling". This provides the double opportunity to conceptualize human-inspired processes of decision-making under uncertainty in the self-adaptive part of a software, as well as to source human unique emotional competences in co-adaptive architectures. In this paper, some algorithms are discussed that can provide software systems with realistic decision-making, and some architectures are conceptualized that resort to human emotions to quantify uncertainty and to contribute in the software's adaptation process.