What “bias in AI” means

A simple explanation of bias in AI and why it matters

Bias in AI shows up as patterns that lead to unfair or uneven outcomes. Those patterns often come from the data a system is trained on, which can reflect existing imbalances.

For example, a hiring tool trained on past recruitment decisions might favour certain types of candidates if those patterns were present in the data. The system isn’t intentionally biased, but it learns from what it sees.

The issue is how that behaviour scales. What might have been inconsistent or subtle in human decision-making can become more consistent, and harder to notice, once it’s automated.

So the challenge isn’t only spotting bias, but understanding where it comes from and how it shows up in decisions.

Which raises a broader question: if AI reflects past behaviour, how do we stop it repeating the same patterns?