Keynote: Explainable AI in Pervasive Healthcare: Open Challenges and Research Directions

Abstract

In our ageing society, there is increasing need for innovative tools to early detect fragility and non communicable diseases of the elderly population. To this aim, in the last few years, several research efforts have been made to exploit sensorized smart-homes and artificial intelligence (AI) methods to detect health issues of the elderly. However, most of these AI systems act as black-boxes, leading to a low level of trust by the clinicians and by the final users. Moreover, they provide limited support to clinicians in making a diagnosis, since they do not provide any explanation of the reason why a given prediction was computed. This talk addresses this challenging problem and our research efforts on this topic. The talk will present a novel AI system to detect cognitive decline symptoms in smart homes, which is supported by explainable AI capabilities. The system relies on clinical indicators of abnormal behaviors, spatial disorientation, and wandering. An AI-fueled dashboard allows clinicians to inspect anomalies together with the explanations of predictions. The system was experimented with a large set of real-world subjects, including people with MCI and people with dementia.