Policy Evaluation of Issue Engagement with Interactive Data Visualizations in Water Security Dashboards

Abstract

Water security represents one of the most critical global challenges of the twenty-first century, necessitating sophisticated decision support systems to manage complex hydrological resources. Interactive data visualizations embedded within water security dashboards have become standard tools for communicating multidimensional environmental data to stakeholders, policymakers, and the general public. However, measuring the true impact of these dashboards remains problematic, as superficial interface interactions do not necessarily correlate with deep, cognitive issue engagement. This paper proposes a novel approach to predicting issue engagement by applying policy evaluation techniques derived from reinforcement learning to user interaction sequences within data visualizations. By conceptualizing user navigation through a dashboard as a sequence of state-action transitions within a Markov Decision Process, we utilize offline policy evaluation methods to estimate the expected value of specific interaction trajectories. This methodology allows for the prediction of subsequent issue engagement without relying on explicit user feedback or interruptive surveys. Through a comprehensive analysis of user interaction logs from a deployed water security dashboard, this research demonstrates that sequential policy evaluation significantly outperforms traditional aggregate statistical models in predicting meaningful engagement outcomes. The findings offer profound implications for the design of adaptive environmental dashboards that dynamically respond to user behavior to maximize comprehension and facilitate informed decision-making in water resource management.