Questions you might want to ask about AI systems
For those not wanting to use the full Toolkit, but just need some questions to ask about AI systems
The Toolkit might be interesting for you even if you are not in Public Health. You might find it useful to know what questions you can ask about an AI system. Or you might want to learn about what other people might ask about an AI system. In this case, this section is for you. Here you can find a list of questions generated by members of the public during the engagement workshops. They cover various areas such as the background of an AI system or how fair it is.
Below you can find a list of questions people might have about an AI system. To see the questions for/under each section, please click the small arrow to expand the field. Alternatively, if you would like to see the full set of resources, click the button below.
List of questions
What’s the decision about?
What is the AI system being used for?
Does the person using the AI system have enough time and training to use it correctly?
What data was used to train the AI system?
Was a broad range of data used?
Was the data relevant to the people affected by this decision?
Is people’s personal information being used? If so, how will it be kept safe?
How well does the AI system work?
How reliable and accurate is the system?
What's the proof for this?
The AI system's background:
Who developed the AI system?
Why did they want to make the AI system?
Do other people recommend the AI system?
Is the AI system recommended by previous users?
Has the AI system been approved by an independent expert body?
How fair and unbiased is the AI system?
Does the system work equally well for all groups of people?
How well does the system take diversity into account?
Has anything unfair ever happened, due to someone using this AI system to help make decisions?
Is there a clear explanation for how the AI system works?
How much do we know about how the AI system comes up with its results?
Human support:
Is there a human who can answer questions about the AI system, if it's giving unclear or strange results?