Validation and Confidence of Socio-technical Models for Infrastructure Resilience

Abstract

Large-scale technical systems, such as critical infrastructures, are embedded in the social environment through the behavior of people using or operating these systems. In case of a disruption, the social environment can be instrumental in maintaining, restoring or, contrarily, decreasing the systems' function. Therefore, it is crucial to consider the social environment in the model-based assessment of a system's performance in the resilience cycle (the system withstanding, absorbing, adapting to, and recovering from disruptions). However, socio-technical models are difficult to validate. To address this difficulty, the presented research investigates three research questions: First, we identify key challenges in validating socio-technical models for resilience assessment. The challenges include the lack of experimental and historical data as well as limitations of traditional statistical methods to address the potentially chaotic nature of complex socio-technical systems. Second, we discuss the extent to which validation and model confidence is achievable for these models. The challenges necessitate approaches beyond classical validation through a data driven proof of correctness. We argue that validation of socio-technical models for resilience should assess model plausibility and quantify the confidence of model outputs. Third, we outline a method implementing the presented validation paradigm. The method uses pattern-based validation to check plausibility of models. Furthermore, Bayesian methods are used for estimating model confidence and integrating new information during all phases of the resilience cycle. The presented approach aims to close the research gap between expectations and practice of validation of socio-technical models for resilient systems.