What you are buying
Choose one decision the system supports and identify who may be affected by it. Review which sources of bias you can examine at each stage and what evidence would help the team decide where a more specific assessment is needed.
This technical report explains where bias can appear in AI systems and in decisions supported by AI. The subject extends beyond training data to the choices people make about objectives, how the system is evaluated and how an output is interpreted in use.
Choose one decision the system supports and identify who may be affected by it. Review which sources of bias you can examine at each stage and what evidence would help the team decide where a more specific assessment is needed.
Yes. It provides context for investigating bias in AI systems and AI-assisted decisions. Consider whether the source is the data, the model, the way the output is used or a change in the population affected.
Choose evaluation methods for the actual decision, groups affected and possible harms. The report helps you frame the investigation and understand why different contexts call for different evidence.