Discover Unknown Biases in AI Systems

Known bias isn't always a problem – a breast cancer model trained mostly on women is rational. The real challenge is finding unknown biases: what aren't we measuring that actually matters?

Use the Johari Window framework to systematically explore where biases might be hiding in your AI system.

Known to Developers
Unknown to Developers
Measurable by Model
Not Measurable by Model
🔍

Arena / Open

Known & Measurable

  • Intentional design choices
  • Documented limitations
  • Accepted trade-offs
Click to explore →
💡

Blind Spot

Unknown but Measurable

  • Missing metrics
  • Unmonitored subgroups
  • Wrong benchmarks
Click to explore →
🔒

Facade / Hidden

Known but Not Measurable

  • Known data gaps
  • Deliberate exclusions
  • Unmeasurable aspects
Click to explore →
❓

Unknown Unknown

Neither Known Nor Measurable

  • Missing metadata
  • Unknown patient groups
  • Untracked outcomes
Click to explore →