The growing complexity of decision-making environments demands intelligent systems that not only assist but also actively enhance human decision-making. This study builds on the Progressive Disclosure Adaptive Questioning Interface (PDAQI) and introduces an advanced framework called the Explainable Reinforcement Learning-based Adaptive Questioning System (ERL-AQS). The proposed system enhances traditional Bayesian adaptive questioning by incorporating reinforcement learning (RL), explainable AI (XAI), and real-time behavioral analytics to refine questioning strategies dynamically through repeated interactions. In contrast to static or short-sighted expected value of information (EVOI)-based systems, ERL-AQS develops optimal questioning techniques through ongoing interaction, balancing exploration and exploitation while reducing cognitive burden. The architecture features belief modeling, policy learning, layers of explainability, and tracking of multi-modal interactions. A simulation-based assessment reveals improved decision-making efficiency, reduced redundancy in questioning, and increased user trust. This research offers a scalable, self-learning decision intelligence framework that can be applied in various fields, including healthcare, financial advisory, governance systems, and digital public infrastructure.
Ganesan et al. (Thu,) studied this question.