Leveraging multiple digital footprint datasets to predict racial, sex-based, and sexual-orientation bias across US states

Derecki, Raphael; O'Shea, Brian; Goulding, James

Abstract

Introduction & Background
Racial, gender, and sexual-orientation biases are pervasive throughout society. Importantly, modern digitally oriented datasets can elucidate important societal variables and potential solutions. One contemporary theory that attempts to explain these biases is parasite-stress: an evolutionary psychology hypothesis suggesting that increased infectious diseases rates increase out-group biases. We present preliminary findings that suggest that disease rates are a meaningful geospatial predictor of multiple biases.


Objectives & Approach
We explored biases using geospatial analyses throughout multiple datasets based on US participants: Project Implicit, American National Election Studies (ANES), Google Trends, and Twitter/X. We included state-based variables to compare between states and assess the most important environmental-level predictors of biases. We built generalised linear and linear mixed-effect models and general linear models. Within Project implicit (n > 3,000,000) and ANES datasets (n > 30,000), we assessed racial and sexual-orientation biases via explicit and implicit measures. For Google Trends and Twitter/X datasets, we assessed racial and sex-based biases via search and tweet-per-state scores. To analyse the biases, we included environmental-level variables, e.g., infectious disease rates (developed by Thornhill and Fincher in 2014), and individual-level variables, e.g., political orientation.


Relevance to Digital Footprints
These preliminary findings analyse everyday people’s online behaviour including volunteered surveys, searches and posts. We attempt to address the pressing societal issue of bias by leveraging modern datasets. Our primary goal is to aid policy makers by recommending cost-effective solutions that can improve several factors of the population’s quality of life.


Results
We find that the most consistently significant predictor of racial bias is infectious disease rates. When leveraging Google Trends data including anti-women terminology, infectious disease rates and population density are consistent predictors of bias. Finally, we find preliminary results suggesting that increased levels of infectious diseases increases homophobic bias.


Conclusions & Implications
Overall, we find that as infectious disease rates increase in a state, the level of racial and sexist bias significantly increases. Consistent with parasite-stress theory, we argue that focusing on reducing infectious disease rates in an area can have a plethora of benefits including improving physical and mental health and reducing biases that damage society.

Citation

Derecki, R., O'Shea, B., & Goulding, J. (2024). Leveraging multiple digital footprint datasets to predict racial, sex-based, and sexual-orientation bias across US states. International Journal of Population Data Science, 9(4), Article 15. https://doi.org/10.23889/ijpds.v9i4.2429

• Journal Article Type: Article
• Acceptance Date: Jun 10, 2024
• Online Publication Date: Jun 10, 2024
• Publication Date: Jun 10, 2024
• Deposit Date: Aug 28, 2024
• Publicly Available Date: Oct 9, 2024
• Journal: International Journal of Population Data Science
• Print ISSN: 2399-4908
• Electronic ISSN: 2399-4908
• Peer Reviewed: Peer Reviewed
• Volume: 9
• Issue: 4
• Article Number: 15
• DOI: https://doi.org/10.23889/ijpds.v9i4.2429
• Public URL: https://nottingham-repository.worktribe.com/output/38905941
• Publisher URL: https://ijpds.org/article/view/2429
• Additional Information: Conference Proceedings for Digital Footprints Conference 2024