The IAT measures group patterns, not individual bias levels

Evidence

However, since recent critiques of the IAT methodology have led some to discount the research linked to implicit bias altogether (Bartlett, 2017), we are careful to explain that the IAT is not akin to a DNA test. It is neither a precise nor an entirely stable measure of bias in any single individual; rather it reveals patterns and tendencies among large groups of people (Godsil et al., 2014). — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1367

The IAT is a computer task that measures how quickly participants can link particular groups with positive or negative words (race attitude) or different racial groups with weapons (weapons association) by pressing a particular key on the computer’s keyboard. We generally explain that the IAT operates much like the Stroop test and can be a useful way for individuals to begin to identify their brain’s associations and, in turn, vulnerabilities to implicit biases. However, since recent critiques of the IAT methodology have led some to discount the research linked to implicit bias altogether (Bartlett, 2017), we are careful to explain that the IAT is not akin to a DNA test. It is neither a precise nor an entirely stable measure of bias in any single individual; rather it reveals patterns and tendencies among large groups of people (Godsil et al., 2014). We find it useful to note that studies on the effects of implicit bias use a wide range of measures—these include functional Magnetic Resonance Imaging (fMRI) (Phelps et al., 2000), facial electromyography (EMG) (Vanman et al., 2004), and cortisol responses (Page-Gould et al., 2008). — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1367

Categorization

• QUESTION: The Bartlett critique that made critics abandon implicit-bias research as a whole — worth digging into what exactly it claimed and whether it holds.
• DEFINE: Clarifies what the IAT does and does not measure — a common misconception corrected.
• XREF: Ties to the reader’s research on professed conviction diverging from measured bias — the IAT’s instability at the individual level is central. — connects to: IAT response latency reveals implicit prejudice even absent self-reported bias — Wade C. Rowatt and Lewis M. Franklin, Christian Orthodoxy, Religious…, 26
• corroborated: 3 independent passages support this claim — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 1367