Recommendations
Below is a summary of the recommendations for clinicians to avoid algorithmic bias in using AI systems for health care.
- Determine who you can seek guidance from in your organization on using AI systems and be sure to ask questions when they arise.
- Consider who was represented in the development of the system and their goals.
- Review how the system was approved and how it is monitored.
- Check the training data, determine how bias could have influenced the data, and if that bias was adjusted for.
- If the training data is not available, advocate to your organization and the developers to know more about the system.
- If the system does not provide a level of certainty, adjust your level of trust in the system output accordingly.
- Think about the harm that could come from using the AI system and how likely that harm is to occur.
- Calibrate your trust in the system based on how much you know about it and the risks associated with using it.
- Determine how the system fits in the overall care of the patient and the larger health care system.