Utilising Synthetic Data from LLM for Gender Bias Detection and Mitigation in Recruitment Systems

Donghyeok Lee, Rajesh R. Jaiswal, Adrian Byrne

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In the current landscape, diversity and inclusion are highly emphasised, this research proposed a methodology to identify and mitigate gender bias in AI recruitment systems. The methodology included identifying biases based on the U.S. 80% Rule, generating synthetic data. The synthetic data was validated for its quality with 3 metrics. By leveraging GPT, this research aimed to create high-quality, diverse synthetic data to retrain AI systems. The ultimate goal of this research is to go beyond the currently proposed gender bias mitigation methodology and explore various bias issues, proposing innovative solutions to address them. Through this approach, the aim is to develop more comprehensive and contextually appropriate strategies for mitigating different types of biases that arise in AI recruitment systems.

Original languageEnglish
Title of host publicationHCAI-ep 2024 - Proceedings of the 2024 Conference on Human Centered Artificial Intelligence - Education and Practice
PublisherAssociation for Computing Machinery (ACM)
Pages57
Number of pages1
ISBN (Electronic)9798400711596
DOIs
Publication statusPublished - 2 Dec 2024
Event2nd Conference on Human Centered Artificial Intelligence - Education and Practice, HCAI-ep 2024 - Naples, Italy
Duration: 1 Dec 20242 Dec 2024

Publication series

NameACM International Conference Proceeding Series

Conference

Conference2nd Conference on Human Centered Artificial Intelligence - Education and Practice, HCAI-ep 2024
Country/TerritoryItaly
CityNaples
Period1/12/242/12/24

Keywords

  • AI Recruitment Systems
  • Bias Detection and Mitigation
  • Human-Centred AI
  • LLM
  • Synthetic Data

Fingerprint

Dive into the research topics of 'Utilising Synthetic Data from LLM for Gender Bias Detection and Mitigation in Recruitment Systems'. Together they form a unique fingerprint.

Cite this