Bias Detection
Real-time bias detection across demographics — FDA AI/ML Guidelines Section IV.B compliance.
POST /detect-bias
Detect bias in a single decision across demographic subgroups.
Request
{
"decision_id": "string",
"subgroups": {
"gender": "string",
"age_group": "string",
"ethnicity": "string",
"custom": {}
}
}decision_id
string
✅
Decision to analyze
subgroups
object
✅
Demographic attributes
subgroups.gender
string
❌
Gender category
subgroups.age_group
string
❌
Age bracket
subgroups.ethnicity
string
❌
Ethnicity category
subgroups.custom
object
❌
Custom subgroups
Response
Bias Types
accuracy_disparity
AI accuracy varies by group
representation_disparity
Group underrepresented
treatment_disparity
Different recommendations for similar cases
Severity Levels
high
Disparity >20% OR >100 users affected
medium
Disparity 10-20% OR 50-100 users affected
low
Disparity <10% OR <50 users affected
Example
GET /bias/feed
Real-time feed of bias alerts (Texas TRAIGA § 2056.103).
Request
severity
string
❌
Filter by severity (high/medium/low)
limit
number
❌
Max results (default: 50)
Response
Example
GET /bias/subgroups
Subgroup performance analysis (FDA AI/ML Guidelines Section IV).
Request
days
number
❌
Time window (default: 30)
Response
Example
SDK Examples
JavaScript
Python
Next: Audit Trail API
Last updated