The ethical relevance of explainibility for AI-solutions for early diagnosis of dementia

The AI-MIND EU funded project aims to develop an artificial intelligence (AI)-based digital tool, which will allow to estimate the\r\nrisk of people with mild cognitive impairment to develop dementia. To facilitate its implementation in clinical practice, in the\r\nproject attention is paid to the Health Technology Assessment (HTA) principles, criteria and model. HTA requires to assess\r\nclinical, safety, economical, organizational, ethical and legal implications of a given health technology (HT).\r\nIn the AI-MIND project, first, a group explored the potential impact of the use of an AI on the doctor-patient relationship, then\r\nHTA experts identified potentially relevant issues for AI, and finally bioethical experts, trained in HTA, conducted an analysis on\r\nthe role of explainability, a new ethical issue to take into consideration in an HTA process regarding AI-based HT.\r\nA Delphy panel conducted ended in April 2023 allowed to identify HTA issues potentially relevant for AI. 46 qualified experts\r\nfrom mainly provided their feedbacks. The highest level of consensus was reached on the importance of ethical issues,\r\nexplainability included.\r\nBioethical experts conducted an analysis on the role of explainability for AI-based digital tools for the risk assessment to\r\ndevelop dementia. Both global (information on the model as a whole) and local explainability (information on a specific\r\nprediction) were investigated and applied to the specific use case. Attention was dedicated to the designated role in the\r\ndecision-making process (algorithm-based, -driven, or –determined) of the AI-MIND predicting tool.\r\nFrom the HTA perspective, explainability is increasingly considered crucial for ethical analysis and should be assessed in\r\nrelation to other aspects such as the design of the technologies, the population target, or the market access.\r\nIt should be one of the topic to address for building and implementing ethical AI-solutions for early diagnosis of dementia.