Definition and scope of algorithmic bias

Summary

Algorithmic bias is the systematic and repeatable tendency of computerized sociotechnical systems to produce “unfair” outcomes—such as privileging one group over another—even when that result may not be the intended function of the algorithm. Bias can be intentional (e.g., biased design choices) or unintentional, arising from how data is coded, collected, selected, and used to train or operate algorithms, as well as from how algorithms are deployed and interpreted in real-world contexts. The scope of algorithmic bias extends beyond purely technical errors. It includes biases embedded in pre-existing cultural, social, or institutional expectations; biases introduced through feature/label choices and data sourcing; and emergent biases that appear when algorithms are used in unanticipated ways, by different audiences, or within feedback loops that reinforce earlier patterns. Because algorithms are often treated as neutral and authoritative, their outputs can displace human judgment and responsibility, potentially amplifying existing inequalities in areas such as search, hiring, healthcare, and criminal justice.

Key Points

• 1

  Algorithmic bias refers to systematic, repeatable unfair outcomes produced by computerized systems, stemming from both intentional design and unintended data/usage factors.

• 2

  Bias can originate from pre-existing societal inequities, data and labeling choices, technical limitations, emergent correlations, unanticipated uses, and feedback loops that reinforce discrimination.