Decision makers will want to realise the potential of AI for genomic medicine by speeding up delivery whilst minimising harms. We outline seven practical policy actions to help achieve this goal.
Artificial intelligence (AI) offers great potential for health in both medical research and care, but the computerised analysis of data by opaque (black box) machine learning processes poses significant practical, ethical and regulatory challenges.
Medical applications of machine learning (a form of AI) offer exciting opportunities for automated data analysis to provide new insights and more accurate predictions than even expert professionals can achieve. This could offer a wealth of benefits, from more efficient triaging and scheduling of appointments and rotas through to earlier detection of disease, more precise diagnosis and treatment, and more effective public health surveillance. However, the logic involved and the processes by which these machine learning models reach conclusions is not obvious – a potential hurdle for effective regulatory oversight, and for securing user and public trust.
The ability to explain how machine learning models work is important to:
- Evidence safety and effectiveness
- Facilitate human-computer interaction
- Assist in scientific or causal understanding
- Underpin control by the data subject or controller accountability
It may be particularly important to secure an adequate explanation where decisions result in high-stakes outcomes, which is often the case in medical research or healthcare.
