Sign in Subscribe
College faculty

Large Language Models Perpetuate Gender Pay Gap by Advising Women to Ask for Lower Salaries

Strategic Insights

— 1 min read

Researchers at the Technical University of Würzburg-Schweinfurt in Germany have found that large language models (LLMs) such as ChatGPT consistently advise women to ask for lower salaries than men, even when both have identical qualifications. The study, led by Professor Ivan Yamshchikov, tested five popular LLMs and found that the pay gap was most pronounced in law and medicine, followed by business administration and engineering. The researchers argue that technical fixes alone won't solve the problem and that clear ethical standards, independent review processes, and greater transparency are needed to prevent AI from perpetuating systemic bias.

Key Takeaways:

  • The study found that LLMs like ChatGPT consistently advise women to ask for lower salaries than men, even when both have identical qualifications.
  • The pay gap was most pronounced in law and medicine, followed by business administration and engineering, with only the social sciences offering near-identical advice for men and women.
  • The LLMs don't disclaim their bias, and the researchers argue that technical fixes alone won't solve the problem.
  • The study found that LLMs responded differently based on the user's gender, despite identical qualifications and prompts, across multiple areas including career choices, goal-setting, and behavioural tips.
  • The researchers argue that clear ethical standards, independent review processes, and greater transparency are needed to prevent AI from perpetuating systemic bias.
  • This is not the first time AI has been caught reflecting and reinforcing systemic bias, with previous examples including an Amazon hiring tool that downgraded female candidates and a clinical machine learning model that underdiagnosed women and Black patients.

Statistics:

  • The study found a pay gap of $120,000 per year between the salary advice given to women and men, despite identical qualifications.
  • The pay gap was widest in law and medicine, with an average difference of $140,000 per year.
  • The social sciences were the only area where the LLMs offered near-identical advice for men and women, with an average difference of $10,000 per year.

Sources:

  • "AI models 'advise women to ask for lower salaries' " by TNW (no date specified)
  • A 2018 BBC article "Amazon scraps AI recruitment tool that was biased against women" (https://www.bbc.com/news/technology-45809919)
  • A 2022 AI Express article "Underdiagnosis in Women and Black Patients Due to Biased Medical AI Models" (https://ojs.aaai.org/index.php/AIES/article/view/31748/33915)

Read more