Overview

Bias is any systematic tendency to prefer one idea, interpretation, group or outcome over others. It appears across many domains — thought, language, journalism, science, technology and measurement — and can be conscious or unconscious. A biased account selects or emphasizes information that supports a particular view while downplaying or omitting countervailing evidence. Understanding bias helps people evaluate claims more carefully and design methods that reduce unfair or incorrect conclusions. For an introductory resource, see related material.

Common forms of bias

Bias is not a single phenomenon; it includes several distinct kinds that are often discussed separately:

  • Cognitive bias: mental shortcuts or patterns of thinking that distort judgement. Examples include confirmation bias (favoring information that confirms existing beliefs) and availability bias (overestimating the importance of information that is easy to recall).
  • Social and cultural bias: stereotypes or preferential treatments rooted in cultural norms, language, or institutional practices. These can influence hiring, policing, education and everyday interactions.
  • Media bias: editorial slants in reporting and commentary. A publication may consistently present a political perspective, selectively quoting sources or framing stories to favor one side; a newspaper, for instance, can show partiality toward a party or ideology as a result of staff views or editorial policy (media bias examples).
  • Statistical and sampling bias: distortions introduced when data are collected or analyzed in ways that do not represent the target population. Selection bias and survivorship bias are common subtypes.
  • Measurement bias: systematic errors in how variables are measured. For instance, measuring someone’s height while they wear shoes adds a predictable extra amount to the result; without reporting shoe height, the measurement is biased and misleading (measurement bias, height example, shoes and measurement).
  • Algorithmic and model bias: when automated systems reflect or amplify historical inequalities or the biases present in their training data, leading to unfair or inaccurate outputs.

Causes and consequences

Causes of bias include limited information, incentives (political, commercial or social), cognitive shortcuts, poor experimental design, and historic injustices encoded in data. Consequences range from minor misunderstandings to serious harms: misleading public debate, unfair treatment of individuals or groups, erroneous scientific conclusions, and flawed policy decisions. In technology, biased models can produce unequal access to services or reinforce discrimination.

Detecting and reducing bias

Detecting bias requires critical scrutiny of methods, language and incentives. Practical steps to reduce bias include:

  • Use representative sampling and transparent measurement protocols in research.
  • Apply blind procedures (where feasible) and preregister analyses to avoid selective reporting.
  • Encourage editorial balance, fact-checking and diversity of sources in journalism.
  • Audit algorithms and datasets for disparate impacts; document training data and model limitations.
  • Promote awareness of common cognitive biases through education and decision aids.

History and notable distinctions

The English word "bias" traces back to Old French biais, meaning a slant or oblique direction; the term came into wider use as a metaphor for tilted judgment. It is important to distinguish bias from related concepts: prejudice generally denotes a preconceived negative attitude toward people or groups; an error describes an unintentional mistake that may be random rather than systematic. Not all biases are strictly harmful: some heuristics speed decision-making in complex environments, though they can still produce systematic errors.

Because bias can be subtle and multidisciplinary, combating it usually requires combining methods: better study design, transparent reporting, diverse voices in institutions, algorithmic oversight, and media literacy among consumers. Recognizing bias is the first step toward more reliable information and fairer outcomes across public life.