The Influence of Instructions to Correct for Bias on Social Judgments

Abstract

We examined how instructions to correct for bias influenced judgments of a male target person whose behavior towards a female was either negative or ambiguous. Half of the female participants with egalitarian or traditional views about gender were instructed to correct for bias prior to reading the vignette. All participants rated his negative behavior unfavorably. In the non-instructed condition, participants with a traditional bias rated the ambiguous male behavior more favorably than participants with an egalitarian bias. However, in the instructed condition, this pattern was reversed. Results demonstrate that the evaluative implications of behavior can impact correction effects.

ACKNOWLEDGMENTS

We thank Marisa Knight and the reviewers for their valuable comments on an earlier version of this article.

Notes

¹Several notable models exist accounting for correction effects in social judgments (e.g., Fazio and Towles-Schwen's, 1999, motivation and opportunity as determinants model; Petty and Wegener's, 1993, 1997, flexible correction model; Martin's, 1986, set/reset model; and Schwarz and Bless's, 1992, inclusion/exclusion model). Although the mental contamination model and flexible correction models are more heavily referenced here, similar predictions can be made with these aforementioned models.

²This study conducted by Stapel and colleagues has not been retracted: http:/c.ymcdn.com/sites/www.spsp.org/resource/resmgr/docs/nonretractionreportpspbinpre.pdf

Note. N = 125 women for all analyses. Scores for Ambivalent Sexism Inventory, Hostile Sexism, Benevolent Sexism ranged from 0 to 5.

³Hierarchical regression analyses were performed using hostile and benevolent subscales as predictors. The main effects of advice type, instruction set, and hostile sexism scores were entered on the first step, all two-way interactions were entered on the second step, and the three-way interaction was entered on the third step. Results revealed a nonsignificant three-way interaction (β = .33, p = .073, R ² = .33), F(7, 114) = 7.95, p < .01. Similar analyses using the Benevolent subscale as the predictor instead revealed the same pattern of results for the three- way interaction (β = .27, p = .104, R ² = .34), F(7, 114) = 8.23, p < .01.

Note. Regression coefficients are reported from the step on which each variable was first entered.

*p < .05. **p < .01.