Glossary

What is explainable AI?

Explainable AI is the set of methods for showing why an AI system produced a particular result. For a traditional scoring model, such as one rating loan applications, it can show which factors raised or lowered the score. For large language models it is much harder, because the answer comes from billions of learned numbers and the model cannot reliably explain itself. Research from AI labs has found that even a reasoning model’s visible working does not always reflect what drove its answer.

Why it matters for a business

Where an AI output feeds a decision about a person, in hiring, credit or insurance, being able to explain it can be a legal requirement as well as good practice. In the EU, data protection law and the AI Act give people affected by significant automated decisions a right to meaningful information about how they were reached. Where explanation matters, a classical model with a readout may be the better tool, or a language model can be limited to reading and summarising while the decision rests on rules or a person. A record of what the AI was given and what it produced makes any decision easier to account for later.