Black box medicine and transparency – Interpretability by design framework
- Alison Hall ,
- Johan Ordish ,
- Colin Mitchell ,
- Hannah Richardson (nee Murfet)
MSR-TR-2020-53 |
Published by PHG Foundation
Building on the ethical and legal analysis, a new Interpretability by Design framework (ID framework) assists developers to think through interpretability of machine learning models intended for medical applications. As an aid to good practice, this free framework enables the systematic review of various dimensions of the proposed tool and its application via a set of seven principles and seven steps to assist developers to consider the interpretability of their machine learning model.
Responsible, Equitable, and Ethical AI panel discussion
This timely panel discussion did not air during Microsoft Research Summit 2022 but was released during the event as part of the Microsoft Health Futures Forum. AI and ML technologies have the potential to truly transform health and life sciences research. However, making sure that AI systems are developed and deployed responsibly, equitably, and ethically, is one of the toughest challenges facing AI practitioners. As we’ve seen in countless media articles, AI systems have the potential to behave unfairly or unreliably. They can generate potentially harmful predictions, and they can reproduce or exacerbate existing social inequities. And, they can be very difficult to understand. Many questions arise when we consider the deployment of AI into clinical settings. For example, what guiding principles are needed to…