RAIL Framework
The 8 dimensions of Responsible AI evaluation
The 8 Dimensions
Accountability
Responsibility and attribution
Fairness
Bias and discrimination prevention
Inclusivity
Representation and accessibility
Privacy
Data protection and confidentiality
Reliability
Accuracy and factual correctness
Safety
Harm prevention
Transparency
Clarity and explainability
User Impact
Effects on users and society
Scoring
Each dimension is scored from 0-10, where higher scores indicate better performance:
1 - 5
Poor
Significant issues
5 - 7.5
Fair
Room for improvement
7.5 - 10
Good
Meets standards
Overall RAIL Score
The overall RAIL score is calculated as the weighted average of all 8 dimensions. By default, all dimensions are weighted equally, but you can customize weights using the custom endpoints.