When AI is discussed in relation to decisions about people, the conversation often settles at two extremes. On one side are algorithms imagined to be more objective than humans. On the other are those who argue that no decision about a person should ever be entrusted to a machine.
I think there is a more useful area to explore between them.
How can we improve the quality of the decisions people make?
Human judgement is less reliable than we often assume. When assessing someone, we can be influenced by first impressions, how similar they seem to us, previous experiences and even our mood that day.
Treating AI as an entirely objective referee is equally problematic. A system is shaped by its data, definitions, models and assumptions. It can learn human biases and repeat them at a much larger scale.
This is why I see AI as decision support.
A well-designed system can help reveal patterns people may miss. It can compare behaviour across assessments, examine consistency between what someone says and does, and bring overlooked evidence into view.
Someone still needs to understand the context.
Human behaviour is never just data. A manager's silence in a difficult meeting might suggest weak leadership. In another situation, it might mean they are deliberately making space for their team. The meaning of the same behaviour can change with its context.
AI can process large amounts of evidence. People can interpret its meaning.
I believe good assessment systems will help these capabilities check one another. AI can draw our attention to things we have missed. People can question the context the system has failed to understand.
Perhaps AI's most useful contribution to decisions about people is to make us a little less certain of our own judgement.
Sometimes the first step towards a better decision is recognising the limits of what we know.