Measuring Situational Bias or Creating Situational Bias?

Click to increase image sizeClick to decrease image size

Notes

¹ What I am calling cultural knowledge is meant to be broad and inclusive and is perhaps better thought of as the cognitive and affective by-products of what Nosek and Hansen (2008) call “cultural experience.”

² Studies find awareness of gendered cultural knowledge, for instance, in even very young child (Ruble et al., 2006). Cross-sectional tests of samples with common cultural knowledge cannot rule out developmental changes in implicit bias unless one equates implicit bias with cultural knowledge awareness.

³ At all relevant times, this IAT compared the groups “European Americans” to “African Americans” or “White people” to “Black people.” Consider also Rae et al. (2015), who found large disparities in average IAT scores for Whites and Blacks at both the national and state levels.

⁴ One way to adjudicate among measures, and to validate the bias-of-the-crowds theory, would be to obtain independent measures of situational bias and look for convergence with the various indirect measures of bias (Payne et al. suggest measuring and manipulating situational conditions that influence implicit bias to examine influences on behavior). To the extent that behavior is deemed a proxy for situational bias, such as Google searching (Rae et al., 2015) or voting behavior within a particular locale, we already have evidence that indirect measures often do not converge with these situational measures. Correlational evidence of the kind Rae et al. (2015) presented (positive correlations between IAT scores and percentage of racially charged Google queries within a state relative to other states) will not suffice when implicit measures are deemed measures of situational bias because that approach would treat even mean scores in the range of the scale denoted neutral or no bias (as with the mean scores for African Americans in Rae et al.) as valid measures of situational bias in states with high frequencies of racially charged searches.

⁵ For example, should we have expected bias conditions in Missouri to have remained constant between 2007 and 2016? Or would we not expect conditions to have gotten worse following the Michael Brown shooting in August 2014? Using the publicly available race IAT database (https://osf.io/52qxl/), I computed average D scores for Missouri in 2007, 2011, 2015, and 2016, which were respectively .339, .325, .306, and 280. If we treat average race IAT scores as a measure of situational anti-Black bias, then we would conclude that bias conditions have actually improved in Missouri since 2014.