TExKG in Health Domain: The Application of Knowledge Graph Based Framework for Explainable Recommendations in the Contexts of Elderly Care, Mental Health, and Emergency Responses

Generating explainable output of intelligent algorithms is subject to several requirements, which originate from the domain of application, involved stakeholders, and the goal of the algorithm. Industrial and health domains differ in multiple aspects, such as the level of criticality, in terms of generating wrong explanations. Those differences challenge explainability approaches to account for the requirements from each of those domains. In this paper, we investigate the ability of the knowledge graph (KG) based explainability framework (TExKG) in accommodating the explanation requirements from both industrial and health domains. We implement the framework in three different use cases in the health domain: elderly care, mental health, and health-emergency responses. We then evaluate the framework based on its ability to accommodate the requirements from these cases, in comparison to its ability of corresponding to requirements from the industrial domain. Our findings highlight the flexibility of the knowledge graph-based framework towards the different domains of application. We trace this flexibility back to the open-box nature of the KG, and the specialized components that the framework offers to integrate domain experts in the explanation generation task.