A Causal Bayesian Network Model for Resolving Complex Wicked Problems

Abstract

Wicked problems are a specific class of complex problems that emerge from complex adaptive systems (CAS) and stakeholder disagreements on the definition and character of these problems and their possible resolution. Attempts at resolving wicked problems through integration and use of formal methods such as ontologies, Bayesian networks (BN), and complex systems dynamic (CSD) models have been attempted recently but wicked problems continue to defy resolution. This paper argues that this is the result of a lack of ontologically precise causal Bayesian models that adequately represent the hierarchical, dynamic, emergent characteristics and multiple perceptions of CAS and their emergent wicked problems. This paper's contribution is the incorporation of complexity systems theory concepts, namely: perspective, granularity and context, as explicit ontological constructs in a high precision ontological causal BN model, the Granular Contextual Perspectives (GCP) causal Bayesian Network model, using Hidden Markov Model (HMM) formalism to address this shortcoming. Using an illustrative example this conceptual paper shows that the (GCP) causal Bayesian Network model performs better than baseline Bayesian Network models at the visual representation, compact and retractable inference, and machine learning of CAS and their emergent wicked problems. The model is useful at supporting the exploration of possible effects of proposed alternative interventions or prototypical design strategies for resolving a given wicked problem.