Interpretability is how well people can understand why a machine learning model gave the answer it did.
In plain words
Some models are easy to read: you can look at their learned numbers and see what matters. Others, like large neural networks, need special tools before anyone can explain them. Explanations help people check a model for problems such as unsafe or unfair behaviour, though researchers still disagree on how to measure interpretability.
A quick example
Imagine a loan model rejects an application. With a simple, interpretable model, the bank can point to the income and missed payments that drove the decision. With a large neural network, it needs extra tools to give a similar explanation.
Where you'll meet it
- Checking a model for unfair or unsafe decisions.
- Explaining why a model gave a particular answer.
Related
Sources
- Docsinterpretability, Google Machine Learning Glossary · read 29 Sept 2026
- PaperTowards A Rigorous Science of Interpretable Machine Learning, Doshi-Velez and Kim (arXiv) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.