Psychology › Social psychology › Social cognition: how we judge people
Implicit bias and the IAT
1998· Finding or experiment
Mixed evidenceSome support; effects smaller or less consistent than once thought.
The idea that people hold automatic associations about groups that they may not endorse, measured by a reaction-time test.
See it on the Psychology map →What it says
In 1998 Anthony Greenwald and colleagues introduced the Implicit Association Test (IAT). People sort words and faces using two keys; if they are faster pairing, say, White faces with 'good' than Black faces with 'good', the speed difference is read as implicit bias. Versions cover race, gender, age and weight. Millions have taken it on the Project Implicit website, and it became a basis for much corporate and police bias training.
Why it matters
It changed public talk about prejudice, suggesting bias can operate without intent. The stakes are high: hiring, policing and courts all debate what IAT scores mean.
The evidence
Status: mixed evidence. Group-level patterns replicate reliably, but an individual's score shifts considerably between sessions, and scores predict actual behavior only weakly.
Open questions
Whether bias training changes how people act; brief interventions shift IAT scores for hours or days, with little evidence of lasting behavior change.
Connections (2)
▸Implicit bias and the IAT became a leading, contested tool in Social cognition: how we judge peopleMixed evidence
The connection
Social cognition studies the fast judgments we make about people. Researchers once relied on what people said about their attitudes, which is easy to shade. The IAT promised to bypass that by timing responses, and it quickly became one of the field's most used tools. Critics argued it was adopted, including for training, faster than the evidence justified.
Who showed it and when
Anthony Greenwald at the University of Washington, Mahzarin Banaji at Yale and later Harvard, and Brian Nosek; critiques from Hart Blanton, Philip Tetlock and colleagues, 2009 and 2013.
How solid is it?
Thousands of published IAT studies. Meta-analyses disagree on its predictive power: a 2009 one found modest links to behavior, a 2013 reanalysis smaller ones, similar to plain questionnaires.
Open questions
Whether IAT scores reveal a person's attitudes or the surrounding culture's associations; one model treats them as a better measure of places than of individuals.
▸Implicit bias and the IAT is pictured as the fast side of System 1 and System 2 (fast and slow thinking)Mixed evidence
The connection
Dual-process models separate fast associations from slow, deliberate beliefs. The IAT measures millisecond sorting speed, System 1 territory, while questionnaires capture what System 2 endorses. That split explains how someone sincerely committed to fairness can still show biased associations, and why bias is thought to matter most under time pressure, as in split-second decisions.
Who showed it and when
Greenwald and Banaji's 1995 paper on implicit social cognition; Bertram Gawronski and Galen Bodenhausen's 2006 model; Joshua Correll's 2002 video-game studies of split-second 'shoot' decisions.
How solid is it?
Implicit and explicit scores correlate only modestly, as a two-process view predicts; but a 2019 meta-analysis found changing implicit scores did not change behavior, weakening the causal story.
Open questions
Whether implicit attitudes are truly unconscious: in 2014 studies, people predicted their own IAT results fairly accurately.