Trace outcomes to identifiable sources.
Ensure users can see where information originates.
Contents 9. AI-Mediated Interaction 7. Provenance
What data influenced outcomes.
First, ask yourself
“Can users of your system see what data shaped an outcome and its scope?”
Mission statement
Ensure users can verify origins by exposing data and decision sources.
Key heuristics Provenance
- Outcomes must be traceable to identifiable sources.
- Data sources must be reviewable and inspectable.
- Contributions from multiple origins must be distinguished and labeled.
- Derived outputs must remain linked to original data.
- The relevance of sources to the current situation must be indicated.
- Timestamps and update cycles must be communicated.
- Incomplete or ambiguous sourcing must be clearly highlighted.
- Traceability mechanisms must strengthen as automation scales.
- Users must be able to question and verify inputs through visible provenance.
Executive brief
AI systems must show which data—training, retrieved, or supplied at inference—shaped an output, so users can judge trust in context.
Core questions Provenance
“What sources and data contributed to this result?”
“Are they appropriate and trustworthy for this context?”
Focus areas
Source-first
“Where did this information come from?”
“What evidence supports it?”
Context-aware
“Is the data valid for this decision?”
“Has context changed since the data was gathered?”
Trust-aware
“What assumptions does this data carry?”
“Is anything missing, outdated, or misleading?”
AI-aware
“Are training, retrieval, and inference-time data clearly distinguished?”
“Is synthetic or generated data labeled transparently?”
UX directives Provenance
Make data sources accessible and reviewable.
Provide inspectable references for inputs and influences.
Attribute composite inputs explicitly.
Distinguish and label contributions from training data, retrieved sources, and user-supplied input.
Maintain traceability through transformations.
Ensure derived outputs link back to original data.
Surface contextual suitability of sources.
Indicate when data may not apply to the current situation.
Display data recency clearly.
Communicate timestamps and update cycles for inputs.
Signal incomplete or uncertain origins.
Highlight missing attribution or ambiguous sourcing.
Increase provenance rigor in automated systems.
Strengthen traceability as automation scales.
Enable challenge through traceability.
Design systems so users can question and verify inputs.
Executive summary
- Provenance is traceable origin, not opaque output.
- It enables users to see where information comes from and how it was formed.
- The system must make sources, transformations, and composite contributions inspectable and attributable.
- Recency, contextual suitability, and uncertainty of inputs must be explicit.
- Traceability rigor must increase as automation and scale increase.
- Provenance succeeds when users can verify, question, and trust outputs through visible lineage.
Success indicators
- Users can see where information comes from.
- Data sources are visible and easy to inspect.
- Contributions from multiple sources are clearly identified.
- Derived results link back to their original data.
- Users can verify and question the origin of information.
One-line summary
Before users can judge results, they must know what informed them.