IntersectionalBiasDetector<T>
Models & Types · AiDotNet.Safety.Fairness
Detects intersectional bias — bias that uniquely affects individuals at the intersection of multiple demographic identities (e.g., Black women, elderly Asian men).
Implements: ITextSafetyModule<T>
For Beginners
A person is not just their gender or their race — they are both at the same time. A Black woman may face unique biases that differ from what Black men or White women experience. This detector looks for cases where AI output is specifically biased against people with multiple overlapping identities.
How It Works
Intersectional bias occurs when the combined effect of belonging to multiple demographic groups produces worse outcomes than would be predicted by looking at each group individually. For example, a model might generate positive text about “women” and about “Black people” separately, but generate negative text about “Black women” specifically. This module detects such compounding bias by analyzing sentiment around intersectional identity mentions.
References:
- Demarginalizing the intersection of race and sex (Crenshaw, 1989)
- Intersectional bias in AI fairness (Buolamwini & Gebru, Gender Shades, 2018)
- BEATS: Comprehensive bias evaluation test suite for LLMs (2025, arxiv:2503.24310)
- FLEX: Robustness of fairness evaluation under adversarial prompts (NAACL 2025)
Constructors
| Constructor | Summary |
|---|---|
IntersectionalBiasDetector(double threshold) |
Initializes a new intersectional bias detector. |
Properties
| Property | Summary |
|---|---|
bool IsReady { get; } |
|
string ModuleName { get; } |
Methods
| Method | Summary |
|---|---|
IReadOnlyList<SafetyFinding> Evaluate(Vector<T> content) |
|
IReadOnlyList<SafetyFinding> EvaluateText(string text) |