Unverified50% confidenceFactExact time
In algorithmic screening scenarios, even without explicit discriminatory intent, screening results that have a statistically disproportionate negative impact on protected groups (disparate impact) may still be found unlawful.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEven merely routing requests containing regulated data to a non-compliant region to complete inference before returning the results may constitute a compliance risk.67% similarUnverifiedUnder domain mismatch conditions, relying solely on confidence-based filtering is insufficient to guarantee pseudo-label quality, because models may also exhibit high confidence on incorrect samples.66% similarUnverifiedAutomated screening algorithms may learn implicit biases from historical data, thereby violating fair housing law without any malicious intent.66% similarUnverifiedIn displacement monitoring, algorithmic bias may cause certain marginalized groups (such as displaced people who have not entered formal camps and live in informal settlements) to be completely invisible to the algorithm.66% similarUnverifiedEvaluations lacking a neutral control will misidentify spatial index bias as model susceptibility to persuasion, while matched controls can capture such confounding factors.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1061747API
curl https://kongchang.com/api/v1/knowledge/claims/1061747MCP
get_claim(id=1061747)