Mehmet EmrahKonyaTR
← Back to glossary

Bias and fairness

Önyargı ve adalet

D5

Bias and fairness concern how systematic patterns, choices, and impacts can produce unequal or unjust outcomes in an AI system and how those outcomes are evaluated in context.

Review status: 2027-02-24

Technical explanation

They require examination of data, design, human decisions, deployment, affected groups, and the value choices behind an assessment.

Conceptual boundaries

Bias is broader than one statistical disparity, and fairness is not a single universally compatible metric.

Provider-neutral example

A hiring-screening team can compare error patterns across relevant groups, inspect data and workflow choices, and document why a selected fairness criterion fits the stated use.

Limitations

A favorable result on one fairness measure does not remove structural, data, human, or deployment sources of bias.

Related concepts

Atomic claims and evidence

  1. 1.1NIST SP 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence
    Source
    NIST SP 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence
    Source role
    Authoritative source
    Exact locator
    Executive Summary, printed p. i, and Section 2.1.2, printed pp. 6-9
    Supported claim
    NIST identifies systemic, computational, and human bias as distinct sources that can arise across AI contexts and lifecycle stages.
    Last verification
    Review due
    Scope limitation
    This taxonomy supports investigation; it does not make every observed disparity proof of unlawful discrimination.
    1.2OECD AI Principles
    Source
    OECD AI Principles
    Source role
    Authoritative source
    Exact locator
    Human rights and democratic values, including fairness and privacy
    Supported claim
    NIST identifies systemic, computational, and human bias as distinct sources that can arise across AI contexts and lifecycle stages.
    Last verification
    Review due
    Scope limitation
    This taxonomy supports investigation; it does not make every observed disparity proof of unlawful discrimination.
  2. 2.1NIST SP 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence
    Source
    NIST SP 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence
    Source role
    Authoritative source
    Exact locator
    Section 3.2.1, printed pp. 30-32
    Supported claim
    NIST notes that fairness is context-dependent and that fairness criteria can involve trade-offs rather than one universally compatible measure.
    Last verification
    Review due
    Scope limitation
    The existence of trade-offs does not excuse avoidable harm or remove the need to justify the chosen evaluation context.
    2.2Regulation (EU) 2024/1689 (Artificial Intelligence Act), consolidated text of 27 July 2026 — D5 instrument slice
    Source
    Regulation (EU) 2024/1689 (Artificial Intelligence Act), consolidated text of 27 July 2026 — D5 instrument slice
    Source role
    Authoritative source
    Exact locator
    Article 10(2)(f)-(g), consolidated text of 27 July 2026
    Supported claim
    NIST notes that fairness is context-dependent and that fairness criteria can involve trade-offs rather than one universally compatible measure.
    Last verification
    Review due
    Scope limitation
    The existence of trade-offs does not excuse avoidable harm or remove the need to justify the chosen evaluation context.