Scope of use indicates whether the AI-application is merely being researched or its use is limited (‘narrow’ or 0% – 30%), it is used regularly but not widely (‘medium’ or 30% – 70%), or whether it is used widely (‘wide’ or 70% – 100%). These indications are based on desk research only. Some applications are being used ‘under the radar’, indicated by the attribute ‘covert’.

Positive impact (blue) involves e.g. improvement of health and safety, better access to and availability of healthcare, better social well-being, better financial status.

Negative impact (red) involves e.g. undesired precedent, impact on human rights, biased outcomes, lack of full accessibility, negative environmental impact.

Positive+negative impact (orange) involves e.g. both great improvement of health and impact on human rights

Technological robustness looks at security, accuracy and reproducibility of the AI-application. Efficacy looks at whether there is a sound problem definition and if the AI-application contributes to solving the problem.

Governance looks at whether there is a legal basis (necessary) for the deployment of the AI-application, and if for example a ‘sunset clause’ exists. It also looks at purpose limitation, i.e. a guarantee that the AI-application is only used to serve a specific purpose. Accountability looks at if there is a clear structure of responsibility.

The special circumstances of this crisis might lead to a trade-off of one requirement against another that could be different than in normal times. ‘Acceptable trade-offs’ looks at whether an AI-application that under normal circumstances would not be accepted, might be accepted under specific conditions given the urgency of this crisis.