Apps

🧪 This platform is in early beta. Features may change and you might encounter bugs. We appreciate your patience!

Algorithmic Bias

Also Known As: ML model bias Statistical discrimination
Aspect ID: algorithmic_bias

Definition

Algorithmic bias is a technical statistical problem that arises when machine learning models encode and perpetuate historical biases present in training data. This occurs through biased training data, proxy variables that correlate with protected characteristics, optimization criteria that weight majority performance, and feedback loops that amplify initial biases over time.

Examples

A recidivism prediction algorithm (COMPAS) was trained on historical arrest data. Because minority defendants were historically arrested at higher rates due to discriminatory policing, the algorithm assigned higher risk scores to minority defendants with similar actual recidivism rates as white defendants.

A resume-screening algorithm trained on a decade of successful hires at a tech company learns to downrank candidates who attended all-women's colleges, because historical hiring patterns reflect past gender discrimination rather than candidate quality. The algorithm encodes and automates that discrimination at scale.

A bank's loan approval model trained on historical lending data systematically assigns lower credit scores to applicants from certain zip codes. Because those zip codes were historically redlined and underserved, the model treats the resulting lack of credit history as a risk signal — perpetuating exclusion under the appearance of objective data-driven decision-making.

Verification Steps
Verification Steps
Binary yes/no questions that an AI must answer to detect a reasoning pattern in a text.
Each of the 452 aspects has verification steps — simple yes/no questions designed to systematically detect whether a pattern appears in a text. For ad hominem: "Does the argument attack a person rather than their claim?" For false dichotomy: "Are only two options presented when more exist?" This ensures consistent, reproducible analysis.

Binary (yes/no) questions an LLM must answer to identify this aspect:

  1. 1

    Was the model trained on historical data that encodes past discriminatory practices?

    Type: binary
  2. 2

    Are model outputs systematically less accurate or more biased for protected subgroups?

    Type: binary
  3. 3

    Were fairness metrics (e.g., demographic parity, equalized odds) evaluated alongside accuracy metrics?

    Type: binary
  4. 4

    Is the claimed neutrality of the algorithm being used to obscure its differential impact?

    Type: binary
Deep Dive
The expandable detail section on each aspect page with examples, psychology, and counter-strategies.
The Deep Dive section provides in-depth information about each aspect: a real-world example showing the pattern in action, an explanation of why it works psychologically, practical advice on how to counter it, alternative names, and links to related aspects.