# Interpretability

Concept · [Building with AI](https://ecosystem.samcarlton.com/ai)

Also called: mechanistic interpretability, explainability, explainable AI, XAI

Research into what is happening inside an [AI model](https://ecosystem.samcarlton.com/model): which internal patterns represent which ideas and how they lead to an output.

**For example:** Interpretability researchers find a pattern inside a model that switches on whenever the text mentions a particular city.

## Connects to (4)

- [Neural network](https://ecosystem.samcarlton.com/neural-network): A kind of software model made of many layers of simple math units connected by weights, loosely inspired by the brain.
- [Latent space](https://ecosystem.samcarlton.com/latent-space): The internal map of numbers where a model places the things it reads.
- [AI safety](https://ecosystem.samcarlton.com/ai-safety): The work of making AI systems behave safely and of measuring how safe they are.
- [Alignment](https://ecosystem.samcarlton.com/alignment): Work to make AI models pursue what people actually intend and behave according to agreed values, such as being honest and refusing harmful…

Map view: https://ecosystem.samcarlton.com/interpretability