Risk Transparency Through Model Card Documentation

Model cards are a fundamental step towards the democratization of digital health technologies (DHT) by disclosing the details of the model, the purpose of the model, and other relevant information to encourage transparency and reporting of machine learning models. During the model development process, a risk analysis is done to identify and mitigate risks that a model possesses. Through the proposed model card that incorporates risk analysis, users are informed of the potential post-mitigation risks associated with the model and their control measures, the context to use the model, the potential biases within the data, and other details about the model, which would not have been known if not given an accompanying model card with risk analysis.