Evaluators¶
This section documents the individual evaluators used in ASHAI's medical response assessment system.
Overview¶
Each evaluator focuses on a specific dimension of medical AI response quality:
- Medical Accuracy - Medical correctness and evidence validation
- Empathy - Empathy and cultural sensitivity
- Precision - Response precision and relevance
- Language Clarity - Language quality and accessibility
- Composite - Combined scoring with weighted aggregation
Composite Scoring¶
The composite evaluator combines all individual evaluators with the following weights:
- Medical Accuracy: 45% - Most important for medical safety
- Precision: 25% - Directness and relevance
- Language Clarity: 20% - Communication quality
- Empathy: 10% - Patient-centered communication
Integration¶
All evaluators are integrated with Weave for comprehensive trace analysis and leaderboard management.