Quick start Hover over labels
This is a rules-based training sandbox (not a statistical fairness score). Use it to compare controls, document risk, and explain bias modes.
- Set the use case (task + stakes). Then move through tabs left-to-right.
- Tune controls (grounding/RAG, eval rigor, HITL, documentation, guardrails).
- Read Outputs: higher Exposure = more risk if used “as-is”. Higher Reliability/Gov = better controls.
- Interpret charts: Radar shows which bias types dominate; Drivers shows what settings are pushing the score (hover bars).
- Save & compare scenarios (e.g., “Loose prompt / no eval” vs “RAG + red-team + HITL”).
Tip: Hover over underlined labels, the ? icons, and the radar axis text for definitions.
This sets the baseline bias profile for the task.
Saved scenarios live in your browser (localStorage).