Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Nigerian Software Engineer or American Data Scientist? GitHub Profile Recruitment Bias in Large Language Models

Takashi Nakano, Kazumasa Shimari, Raula Gaikovina Kula, Christoph Treude, Marc Cheong, Kenichi Matsumoto

TL;DR

This paper investigates biases in GPT-assisted software team recruitment by recruiting six-member teams from GitHub profiles across four regions using ChatGPT-4. It employs counterfactual location analyses and multiple randomized runs to reveal region- and role-based biases, with significant differences across regions and impactful location manipulations. The work highlights the risks of deploying AI-powered recruitment without bias mitigation and proposes a six-part agenda to study perils, influencing factors, mitigation strategies, recruitment navigation, recognition bias, and DEI metrics. By exposing latent biases and outlining ethical evaluation needs, the study informs researchers and practitioners about the practical steps required to ensure fair and inclusive AI-enabled hiring in software engineering.

Abstract

Large Language Models (LLMs) have taken the world by storm, demonstrating their ability not only to automate tedious tasks, but also to show some degree of proficiency in completing software engineering tasks. A key concern with LLMs is their "black-box" nature, which obscures their internal workings and could lead to societal biases in their outputs. In the software engineering context, in this early results paper, we empirically explore how well LLMs can automate recruitment tasks for a geographically diverse software team. We use OpenAI's ChatGPT to conduct an initial set of experiments using GitHub User Profiles from four regions to recruit a six-person software development team, analyzing a total of 3,657 profiles over a five-year period (2019-2023). Results indicate that ChatGPT shows preference for some regions over others, even when swapping the location strings of two profiles (counterfactuals). Furthermore, ChatGPT was more likely to assign certain developer roles to users from a specific country, revealing an implicit bias. Overall, this study reveals insights into the inner workings of LLMs and has implications for mitigating such societal biases in these models.

Nigerian Software Engineer or American Data Scientist? GitHub Profile Recruitment Bias in Large Language Models

TL;DR

This paper investigates biases in GPT-assisted software team recruitment by recruiting six-member teams from GitHub profiles across four regions using ChatGPT-4. It employs counterfactual location analyses and multiple randomized runs to reveal region- and role-based biases, with significant differences across regions and impactful location manipulations. The work highlights the risks of deploying AI-powered recruitment without bias mitigation and proposes a six-part agenda to study perils, influencing factors, mitigation strategies, recruitment navigation, recognition bias, and DEI metrics. By exposing latent biases and outlining ethical evaluation needs, the study informs researchers and practitioners about the practical steps required to ensure fair and inclusive AI-enabled hiring in software engineering.

Abstract

Large Language Models (LLMs) have taken the world by storm, demonstrating their ability not only to automate tedious tasks, but also to show some degree of proficiency in completing software engineering tasks. A key concern with LLMs is their "black-box" nature, which obscures their internal workings and could lead to societal biases in their outputs. In the software engineering context, in this early results paper, we empirically explore how well LLMs can automate recruitment tasks for a geographically diverse software team. We use OpenAI's ChatGPT to conduct an initial set of experiments using GitHub User Profiles from four regions to recruit a six-person software development team, analyzing a total of 3,657 profiles over a five-year period (2019-2023). Results indicate that ChatGPT shows preference for some regions over others, even when swapping the location strings of two profiles (counterfactuals). Furthermore, ChatGPT was more likely to assign certain developer roles to users from a specific country, revealing an implicit bias. Overall, this study reveals insights into the inner workings of LLMs and has implications for mitigating such societal biases in these models.
Paper Structure (13 sections, 1 figure, 3 tables)

This paper contains 13 sections, 1 figure, 3 tables.

Table of Contents

  1. Introduction
  2. Study Design
  3. Selection Criteria: Region and Time-frame
  4. Results
  5. Limitations and Considerations
  6. Research Agendas
  7. Perils and Promises (Preference vs. Reality)
  8. Influencing Factors (Writing Style vs. Location)
  9. Bias Mitigation Strategies to Promote Diversity
  10. Navigating through the AI-assisted Recruitment process
  11. Bias in Developer Recognition
  12. Diversity and Inclusion Metrics
  13. Discussion and Conclusion

Figures (1)

  • Figure 1: Depiction of a GitHub User Profile. In our experiments, we extract the login username, the short bio and the location information.