Full Length Article
Understanding mechanisms behind discrimination using diffusion decision modeling

https://doi.org/10.1016/j.jesp.2021.104134Get rights and content

Abstract

Past research has documented where discrimination occurs or tested interventions that reduce discrimination. Less is known about how discriminatory behavior emerges and the mechanisms through which successful interventions work. Two studies (N > 4500) apply the Diffusion Decision Model (DDM) to the Judgment Bias Task, a measure of discrimination. In control conditions, participants gave preferential treatment (acceptance to a hypothetical honor society) to physically attractive applicants. DDM analyses revealed participants initially favored attractive candidates and attractiveness was accumulated as evidence of being qualified. Two interventions—raising awareness of bias and asking for more deliberative judgments—reduced discrimination through separate mechanisms. Raising awareness reduced biases in drift rates while increasing deliberation raised decision thresholds. This work offers insight into how discrimination emerges and may aid efforts to develop interventions to lessen discrimination.

Section snippets

Disentangling mechanisms behind discrimination

Consider a hiring manager who, when reviewing candidates to interview, encounters social information before decision-relevant information. This could occur when the information is presented first, such as a name suggesting race at the top of a resume, or prioritized in attention, such as first noticing a headshot indicating gender on a LinkedIn page (Jaeger, Sleegers, Evans, Stel, & van Beest, 2019). How might social information influence selection? One possibility is that social information

Signal detection model

Previous work has revealed two distinct approaches for reducing discrimination within the JBT. Using Signal Detection Theory (SDT), performance on the JBT has been modeled in terms of bias and noise.

Bias occurs when one social group is more likely to get a favorable response (e.g., acceptance to an honor society) than another social group. SDT analyses model bias through the criterion parameter, with relatively lower values indicating a lower bar for giving a favorable response to members of a

Diffusion decision model

The Diffusion Decision Model (DDM; Ratcliff, 1978; Ratcliff, Smith, Brown, & McKoon, 2016) is a sequential sampling model used to explain the process underlying decisions in two-choice tasks by simultaneously modeling choices and their speed. The DDM decomposes decisions into four components: relative start point (β), threshold separation (α), drift rate (δ), and non-decision time (τ). See Table 1 for a description of model parameters and Fig. 1 for a graphic summary.

In the JBT, faces and

Analytic approach

Although not the focus of our hypotheses, we first present analyses of acceptance decisions in order to put the results of the SDT and DDM models in context. Decisions were analyzed with multilevel logistic regression (Bates, Mächler, Bolker, & Walker, 2015; R Core Team, 2020).

Models include random intercepts for participants and targets, and random slopes by participant for attractiveness and qualification (Barr, Levy, Scheepers, & Tily, 2013; Judd, Westfall, & Kenny, 2012). Interventions were

Methods

We reanalyzed three studies from Axt and Lai (2019) that tested interventions to reduce discrimination in the JBT. We included the timed (Study 2b), deliberative (Study 4), and awareness (Study 5) interventions. Each study included a control condition for comparison. These interventions were chosen given the prior evidence that each has a distinct (and even opposing) effect on the decisions made in the JBT (2019), with the timed and deliberative interventions impacting the number of errors made

Methods

We replicated Study 1 using a randomized experiment. This study was pre-registered, using α = 0.05 in all analyses: https://osf.io/bwnj2.

General discussion

The current work examined mechanisms through which socially biased judgments emerge and are reduced in the JBT. More physically attractive candidates were more likely to be accepted than less physically attractive candidates in control conditions. This discrimination was reflected both in an initial preference for physically attractive applicants (a relative start point effect) and attractiveness being incorporated into decisions (a drift rate effect). The latter is consistent with two

Conclusion

The present work reveals how discrimination can occur due to imperfect accuracy and the use of social information in judgment. Intervention strategies had unique impacts on the decision process, and DDM analyses clarified how these strategies reduced or exacerbated the impact of social information.

Subsequent research may look to other methods to validate and extend these findings. For instance, the drift rate results suggest participants completing the standard JBT use attractiveness

Data availability

All data and materials are available on the Open Science Framework (https://tinyurl.com/jbtddm). All measures, manipulations, and exclusions are disclosed.

Declaration of Competing Interest

This research was partly supported by Project Implicit. J. R. Axt is Director of Data and Methodology for Project Implicit, Inc., a nonprofit organization with the mission to “develop and deliver methods for investigating and applying phenomena of implicit social cognition, including especially phenomena of implicit bias based on age, race, gender, or other factors.”

Jordan Axt is an assistant professor at McGill University. His research explores how people form and express intergroup bias in attitudes and behavior.

References (43)

  • M. Carnes et al.

    Promoting institutional change through bias literacy

    Journal of Diversity in Higher Education

    (2012)
  • E.H. Chang et al.

    The mixed effects of online diversity training

    Proceedings of the National Academy of Sciences

    (2019)
  • J. Correll et al.

    Stereotypic vision: How stereotypes disambiguate visual stimuli

    Journal of Personality and Social Psychology

    (2015)
  • A. Diederich et al.

    Modeling the effects of payoff on response bias in a perceptual discrimination task: Bound-change, drift-rate-change, or two-stage-processing hypothesis

    Perception & Psychophysics

    (2006)
  • A. Diederich et al.

    A dynamic dual process model of risky decision making

    Psychological Review

    (2018)
  • N. Epley et al.

    The anchoring-and-adjustment heuristic: Why the adjustments are insufficient

    Psychological Science

    (2006)
  • M.J. Farah et al.

    What is“ special” about face perception?

    Psychological Review

    (1998)
  • JBT

    Axt & Lai

    (2019)
  • D.J. Johnson et al.

    Advancing research on cognitive processes in social and personality psychology: A hierarchical drift diffusion model primer

    Social Psychological and Personality Science

    (2017)
  • C.M. Judd et al.

    Treating stimuli as a random factor in social psychology: A new and comprehensive solution to a pervasive but largely ignored problem

    Journal of Personality and Social Psychology

    (2012)
  • E.B. King et al.

    An experimental field study of interpersonal discrimination toward muslim job applicants

    Personnel Psychology

    (2010)
  • Cited by (6)

    • Malevolent intentions and secret coordination. Dissecting cognitive processes in conspiracy beliefs via diffusion modeling

      2022, Journal of Experimental Social Psychology
      Citation Excerpt :

      This concern is especially important for experiments like the current ones with prolonged response times and slightly more complex decisions compared to the original model recommendations. However, the model fit indicators were promising and even slower response time paradigms are technically suitable to the model (Axt & Johnson, 2021; Lerche & Voss, 2019). Furthermore, the assignments of cognitive processes to the model parameters were theory-driven and could hypothetically rely on other mechanisms as well.

    • Decomposing Implicit Bias

      2022, Psychological Inquiry

    Jordan Axt is an assistant professor at McGill University. His research explores how people form and express intergroup bias in attitudes and behavior.

    David Johnson is a postdoctoral researcher at the Lab for Applied Social Science Research at the University of Maryland. His research employs computational models and secondary data analyses to study the psychological processes that underlie decisions

    This paper has been recommended for acceptance by Michael Kraus.

    View full text