Bias

Bias is a tendency, inclination, or prejudice toward or against something, often manifested in a way that is unfair or unreasoned. In everyday use, bias can refer to personal preferences or to systematic distortions in data or analysis. Distinguishing between informal preferences and systematic bias is important in research and decision making. Bias can arise from cognitive limits, social influences, incentives, or incomplete information.

Cognitive biases are mental shortcuts that help process information but can lead to errors. Common examples include confirmation bias (favoring information that confirms beliefs), anchoring (relying too heavily on the first piece of information), availability (overestimating frequent events), and representativeness (judging based on how much something resembles a stereotype).

Social and institutional biases involve prejudices, stereotypes, or discriminatory practices that affect groups defined by race, gender, age, or other attributes. These biases can be overt or implicit, and may influence hiring, evaluation, or policy.

Data- and measurement-related biases arise in gathering and analyzing information. Sampling bias occurs when the sample is not representative; measurement bias occurs when data collection methods favor certain outcomes; reporting and publication biases can skew what is observed.

Algorithmic bias occurs when computer systems reflect or amplify human biases through training data, model choices, or deployment contexts. This can produce unfair outcomes in hiring, lending, surveillance, or legal settings.

Mitigation involves awareness, diverse perspectives, rigorous methodology, preregistration, blind evaluation where feasible, auditing for fairness, and ongoing monitoring. No method guarantees elimination of all bias, but systematic efforts reduce unfair or inaccurate judgments.