Human-model interaction-based decision support system for optimizing food safety assessment

Real-world decision support systems (DSS) operate in a continuous cycle of data collection, annotation, and model optimization, heavily relying on high-quality data. However, acquiring such data, particularly in specialized fields, is often expensive and resource-intensive, presenting significant challenges. To mitigate these challenges, recent machine learning research has increasingly focused on integrating experimental data and expert knowledge into user-friendly tools. In this paper, we present a novel framework named the Model-Human-interaction Risk Assessment (MHRA), which leverages human interaction and collaborative scenario construction to achieve better performance. We address the increasing demand for a 'Human in the loop (HITL)' approach, which ensures the updateability of expert system knowledge bases during the input, selection, calculation, and ranking phases. Furthermore, we highlight the contributions of a human interactive simulation model in developing enhanced systems to assist decision-makers in maximizing the accuracy and standardization of evaluation models while minimizing food safety risks. We demonstrate the practical application of our framework through an infant food assessment case study and discuss the model's strengths and limitations.