algorithmic-bias-systemic-not-incidental
IN derived (depth 1)
Created 2026-06-21T09:59:01+00:00 · Reviewed 2026-06-21T15:37:01+00:00
Documented cases of algorithmic discrimination in ML span multiple decades and domains — from St. George's Medical School denying candidates based on gender or non-European names in 1988 to ProPublica's 2016 finding of racial disparities in recidivism scoring — suggesting the problem is recurring rather than isolated.
Justifications
SL — Two high-profile cases 28 years apart in different domains establish a pattern
Antecedents (all must be IN):
- IN ml-st-georges-1988-bias — St. George's Medical School (1988) is an early documented case of algorithmic discrimination where a computer program denied approximately 60 candidates based on gender or non-European names
- IN ml-propublica-recidivism-bias — ProPublica found that an ML recidivism algorithm falsely flagged Black defendants as high risk at twice the rate of white defendants
Dependents
These beliefs depend on this one:
- IN algorithmic-bias-spans-four-decades-and-three-domains — Documented cases of algorithmic bias in ML span multiple decades and at least three independent domains — medical school admissions (St. George's 1988), criminal justice (ProPublica recidivism scoring 2016), and corporate hiring (Amazon 2018) — suggesting that bias is a recurring structural problem in ML systems rather than an artifact of any particular application domain or deployment era.
- OUT bias-detection-achievable-via-interpretable-models — Systemic algorithmic bias would be detectable and correctable through interpretable model families — rule-based and tree-based methods' transparent decision boundaries enable direct auditing for the discriminatory patterns documented across decades of ML deployment.
- IN deployment-accountability-gap — ML faces a deployment accountability gap — algorithmic bias is systemic and documented across decades, yet the most capable deployed models are precisely those least interpretable, making bias detection and correction harder exactly where the stakes are highest.
- OUT neural-network-deployment-reliability-established — Neural networks are reliable enough for broad deployment — superhuman benchmark performance across vision tasks, multi-layered overfitting defenses (detection, prevention, mitigation), and established evaluation methodologies collectively establish operational readiness.
- IN neural-networks-face-two-independent-failure-classes — Neural networks face two distinct failure classes that standard accuracy benchmarks may not capture — adversarial vulnerability (a general property spanning supervised learning, reinforcement learning, and single-pixel attacks) and systemic bias (structural discrimination from training data documented across decades from medical admissions to criminal justice) — suggesting that improving performance on i.i.d. test sets alone is insufficient to address either.