Enable triple-loop learning: integrating soft information and human-machine interaction in data-driven decision-making

This research presents a methodology that integrates an organizational learning approach for developing data-driven decision-making (DDDM) to successfully integrate mutual human and machine learning (triple-loop learning). Validated in a non-life insurance case study, the methodology addresses the ineffectiveness of an existing DDDM tool due to the lack of human-machine interaction. The existing DDDM tool is further developed by triple-loop learning. The case study demonstrates how triple-loop learning enables the DDDM tool, consisting of predictive models, to implement human decision norms and values, and enables people to gain new insights from working with the DDDM tool. The research contributes to design science theory by offering a methodology and guidelines to enable triple-loop learning in the development of a predictive model for DDDM in general. Within DDDM, triple-loop learning should lead to the alignment of human and machine mental models so that decisions made by DDDM align with human norms and goals.