> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.umnai.com/concepts/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.umnai.com/_mcp/server.

# Concepts

## Docs

- [Concepts](https://docs.umnai.com/concepts/overview.md): Understand the modelling and explanation concepts behind Umnai.
- [Explainable Neural Net (XNN)](https://docs.umnai.com/concepts/model-concepts/explainable-neural-net-xnn.md): Understand the predictive model structure behind Umnai explanations.
- [Explanation Structure Model (ESM)](https://docs.umnai.com/concepts/model-concepts/explanation-structure-model-esm.md): Understand how Umnai organises model outputs into structured explanations.
- [Induction](https://docs.umnai.com/concepts/model-concepts/induction.md): Understand how the platform creates model structure from training data.
- [Neuro-symbolic AI and hypergraph models](https://docs.umnai.com/concepts/model-concepts/neuro-symbolic-ai-and-hypergraph-models.md): Understand the modelling approach behind XNNs and ESMs.
- [Symbolic reasoning](https://docs.umnai.com/concepts/model-concepts/symbolic-reasoning.md): Understand how explicit concepts, rules, and reasoning strategies support explanations.
- [Explanations](https://docs.umnai.com/concepts/explanation-concepts/overview.md): Understand the main types of explanations produced by Umnai models.
- [Global explanations](https://docs.umnai.com/concepts/explanation-concepts/global-explanations/overview.md): Understand model behaviour across a dataset or segment.
- [Dataset](https://docs.umnai.com/concepts/explanation-concepts/global-explanations/dataset.md): Understand the dataset concepts used in global explanations.
- [Module and feature importance](https://docs.umnai.com/concepts/explanation-concepts/global-explanations/module-and-feature-importance.md): Understand which modules, features, and interactions matter most across a dataset.
- [Module dependency plot](https://docs.umnai.com/concepts/explanation-concepts/global-explanations/module-dependency-plot.md): Understand how a module's attribution changes across its input space.
- [Strengths and weaknesses](https://docs.umnai.com/concepts/explanation-concepts/global-explanations/strengths-and-weaknesses.md): Understand model confidence through strength and uncertainty.
- [Local explanations](https://docs.umnai.com/concepts/explanation-concepts/local-explanations/overview.md): Understand why a model produced a specific prediction.
- [Predictions and attributions](https://docs.umnai.com/concepts/explanation-concepts/local-explanations/predictions-and-attributions.md): Understand how predictions are built from module and feature contributions.
- [Decision action table](https://docs.umnai.com/concepts/explanation-concepts/local-explanations/decision-action-table.md): Understand rules, activations, and attributions in a single local explanation table.
- [Strengths and weaknesses](https://docs.umnai.com/concepts/explanation-concepts/local-explanations/strengths-and-weaknesses.md): Understand confidence-related signals for an individual prediction.
- [Context explanations](https://docs.umnai.com/concepts/explanation-concepts/context-explanations.md): Understand how comparisons and scenarios enrich local explanations.
- [Explanation workflows](https://docs.umnai.com/concepts/explanation-workflows/overview.md): Understand how explanation components support practical analysis and decisions.
- [Model behaviour](https://docs.umnai.com/concepts/explanation-workflows/model-behaviour.md): Use rules and module dependency plots to inspect how a model behaves.
- [Identifying bias](https://docs.umnai.com/concepts/explanation-workflows/identifying-bias.md): Use control swaps, attributions, and MDPs to inspect bias in predictions and models.
- [Justifying attribution explanations](https://docs.umnai.com/concepts/explanation-workflows/justifying-attribution-explanations.md): Use activated rules to explain why an attribution was produced.
- [Identifying data segments](https://docs.umnai.com/concepts/explanation-workflows/identifying-data-segments.md): Use rule conditions to understand the behavioural segments discovered by induction.
- [Identifying and addressing low data coverage](https://docs.umnai.com/concepts/explanation-workflows/identifying-and-addressing-low-data-coverage.md): Use histograms and activations to find sparse regions and decide how to handle them.
- [Model monitoring](https://docs.umnai.com/concepts/explanation-workflows/model-monitoring.md): Use activation data to detect drift, deviations, and unusual deployed-model behaviour.
- [Model optimisation](https://docs.umnai.com/concepts/explanation-workflows/model-optimisation.md): Use explanations to guide feature selection and module creation.
- [Real-time explanations](https://docs.umnai.com/concepts/explanation-workflows/real-time-explanations.md): Understand how predictions and explanations are returned together at inference time.
- [Audit logs](https://docs.umnai.com/concepts/explanation-workflows/audit-logs.md): Use explanation results as records for transparency, accountability, and reconstruction.
- [Decision Space](https://docs.umnai.com/concepts/decision-space.md): Turn predictions and explanations into fit-for-purpose decisions.