Abstract

People are increasingly employing artificial intelligence as the basis for decision-support systems (DSSs) to assist them in making well-informed decisions. Adoption of DSS is challenging when such systems lack support, or evidence, for justifying their recommendations. DSSs are widely applied in the medical domain, due to the complexity of the domain and the sheer volume of data that render manual processing difficult. This paper proposes a metalevel argumentation-based decision-support system that can reason with heterogeneous data (e.g. body measurements, electronic health records, clinical guidelines), while incorporating the preferences of the human beneficiaries of those decisions. The system constructs template-based explanations for the recommendations that it makes. The proposed framework has been implemented in a system to support stroke patients and its functionality has been tested in a pilot study. User feedback shows that the system can run effectively over an extended period.