Neuro-Symbolic Artificial Intelligence for Safety Engineering

Laura Carnevali, Marco Lippi

Abstract: One of the challenges for safety engineering in the coming years is to fully exploit the potential of artificial intelligence technologies, while maintaining a high level of interpretability, trustworthiness, and accountability. In this position paper, we argue that neuro-symbolic artificial intelligence approaches could represent the ideal framework to address tasks in the field of safety engineering, by combining the capability of neural networks of processing large volumes of data and of handling uncertainty, with the expressive power of symbolic approaches in modeling domain knowledge, rules, and constraints.