Carissa Véliz warns AI predictions can undermine democratic choice

Carissa Véliz says AI predictions should not be treated as facts during a speech at a forum in Spain.

The Oxford philosopher argues that democracy depends on an uncertain, open future, because election results are not settled until people vote and ballots are counted.

Machine-learning systems use large amounts of data to find patterns and estimate what may happen, but a prediction is not knowledge of the future.

Véliz says the risk grows when predictions guide decisions about loans, jobs, benefits or court cases.

If a system denies someone an opportunity, it can be hard to understand or challenge the reasoning, and repeated decisions can help make the prediction come true.

She also warns that gathering and concentrating data creates security risks, including theft, leaks or use for other purposes; no next steps are specified.