A risk-based model for human-artificial intelligence conflict resolution in process systems

Abstract

The conflicts stemming from discrepancies between human and artificial intelligence (AI) in observation, interpretation, and action have gained attention. Recent publications highlight the seriousness of the concern stemming from conflict and models to identify and assess the conflict risk. No work has been reported on systematically studying how to resolve human and artificial intelligence conflicts. This paper presents a novel approach to model the resolution strategies of human-AI conflicts. This approach reinterprets the conventional human conflict resolution mechanisms within AI. The study proposes a unique mathematical model to quantify conflict risks and delineate effective resolution strategies to minimize conflict risk. The proposed approach and mode are applied to control a two-phase separator system, a major component of a processing facility. The proposed approach promotes the development of robust AI systems with enhanced real-time responses to human inputs. It provides a platform to foster human-AI collaborative engagement and a mechanism of intelligence augmentation.