Designing ethical AI systems starts with data quality and fairness. Biased datasets can cause AI models to replicate or amplify social inequalities, often without users realizing it.

This research initiative was created to address these risks by making bias detection and mitigation understandable, transparent, and usable by non-expert users.

The project addressed several key challenges in responsible AI development:

  • AI systems may replicate or amplify existing social biases
  • Dataset bias often goes undetected by non technical users
  • Existing tools are complex or lack transparency
  • The European AI Act requires appropriate measures to detect, prevent, and mitigate bias

This called for a solution able to translate complex fairness concepts into practical, user-friendly actions.

BitBang designed the user interface of a research experimental environment to simplify bias assessment and mitigation.

The platform enables users to:

  • Upload their own datasets or explore sample datasets
  • Flag sensitive attributes and potential proxies
  • Apply fairness metrics and bias tests
  • Generate synthetic data to rebalance datasets
  • Interact through an intuitive interface suitable for non technical profiles

The UI plays a key role in guiding users through interpretation and mitigation steps without requiring advanced data science skills.

  • Easier access to fairness and bias analysis tools
  • Support for fair by design AI development practices
  • Increased transparency and ethical awareness
  • Privacy preserving local testing capabilities
  • Planned integration into the EU AI on Demand platform
  • Availability as a standalone tool for wider adoption