⚖️
Algorithmic Bias
Privacy
Systematic error or unfair disparity introduced by data, modeling, or deployment choices.
Definition
Algorithmic bias arises from training data imbalance, proxy variables, feedback loops, and design decisions. Bias can appear as disparate impact, unequal error rates, or differential treatment.
In plain English
Systematic error or unfair disparity introduced by data, modeling, or deployment choices.
Why this matters
Why it matters: Biased profiling and automated decisions amplify privacy harms and discrimination.
Example
Example: Perform fairness testing, document features used, and avoid sensitive proxies unless strictly necessary and justified.