Human-Rights Due Diligence in Medical Artificial Intelligence: Presented at the 6th Global Business and Human Rights Conference, Monterrey 3-4 Sept 2020

Publikation: Working paperForskningfagfællebedømt

Artificial intelligence has spread in all sectors, including in health care. The use of algorithms, in particular machine learning, contributes to find better treatments to patients, and offers promising perspectives to fight cancer for instance. Yet, algorithms are not a neutral health product since they are programmed by humans, with the risk to propagate human rights infringements. Thus, stakeholders should be careful not being too deferent to the algorithms they use. In fact, scholarship identified several human rights risks, including in the medical sector. In order to mitigate those risks, scholars and institutions prepared several reports and guidelines on ethics for medical AI. Human Rights Impact assessments are rarer, especially in the medical area. Yet, a Human Rights approach of medical AI is both necessary on substance and from an organizational and procedural perspective. On substance, a human right approach allows to rely on legal standards rather than unclear notions of fairness for instance. The interrelation between human rights, especially the right to health and the right to be free from discrimination allows to endorse a broader and more consistent approach of AI. From a procedural perspective, the UN guiding principles would allow to target all stakeholders, including the corporations developing health care algorithms. Such approach would establish a chain of duties and responsibilities bringing more transparency and consistency in the overall process. Although this approach would not solve all AI challenges, it would offer a method, a frame for discussion will all relevant actors, including vulnerable populations. The creation of a human rights culture in the techno-science space will benefit from the human rights education of medical doctors and data scientists, and from more collaboration at the initial stages of algorithms making. Building a human rights compliant medical AI will not only rely on checking boxes on legal standards but on education and discussion with all relevant parties and experts.
OriginalsprogEngelsk
StatusIkke-udgivet - 4 sep. 2020

ID: 248333315