Concept Bottleneck Model
A neural network architecture that first predicts human-understandable concepts from the input, then uses those concepts to make the final prediction, enabling interpretability.
In plain language
An AI that first identifies understandable features (such as whether an animal has stripes and has four legs) before making its final prediction (such as it is a zebra). This design makes it easier to understand why the AI reached its conclusion.
Why this matters
Concept bottleneck models improve the interpretability and auditability of AI decisions. For organisations required to explain or defend AI recommendations to regulators or affected individuals, this architecture provides a clearer audit trail and easier detection of conceptual errors.
Relevance
ImplementationConcept bottleneck models are an architectural choice that improves interpretability, supporting better governance oversight of AI decision-making.
Related terms
Putting concept bottleneck model into practice in your organisation?
Ready to transform your AI strategy?
Partner with Australia's AI strategy and governance specialists. From adoption roadmaps to ISO 42001 audit readiness.