Improving the Predictability of Racial Preference Attitudes: Using Machine Learning Models to Predict Concurrent Implicit-Explicit Black-Preference Attitudes
Open Access
Author:
Rodriguez, Raphael
Graduate Program:
Information Sciences and Technology
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
May 21, 2020
Committee Members:
Frank Edward Ritter, Thesis Advisor/Co-Advisor Daniel Susser, Committee Member Shomir Wilson, Committee Member Mary Beth Rosson, Program Head/Chair
Keywords:
racism machine learning implicit iat
Abstract:
The goal of this exploratory paper is to answer the plea for applying novel machine learning techniques to big data in social psychology by a) establishing a more predictable variable of attitudes toward Black people, and b) creating a single model that integrates five previously established stereotype and prejudicial measures. From the publicly available data from Project Implicit (projectimplicit.org), I used the Implicit Association Test (IAT) on Race from 2018—this dataset included a user’s implicit preference for White or Black people, demographic information, and self-reported racial attitudes. As a result, the analysis suggests that demographic and self-reported racial attitudes are better able to predict concurrent implicit-explicit Black-preference attitudes (CIEBA) as compared to implicit or explicit attitudes separately (Avg. R2, MSE: CIEBA = .365, 0.388; explicit-only = .324, 0.672; implicit-only = .110, 0.876). Secondly, self-reported questions that matched prior prejudicial measures (e.g., social distance, victim-blaming, egalitarianism, intergroup anxiety, blatant negative stereotypes) were grouped appropriately and were able to validly predict racial attitudes (R2= .458, MSE = 0.325). This analysis also identified critical thresholds within the stereotype and prejudice measures at which point racial preferences for White people over Black people significantly accelerates.