Archive · 31 Jul 2025
Study: ChatGPT Shows Gender Bias in Salary Advice
Imagine you ask a robot, like ChatGPT, for help with your salary.
A new study found these robots might give different advice depending on if you're a man or a woman.
They sometimes suggest women should accept less money.
Researchers looked at different robots and saw they often did this.
This means the robots might be biased, like they're learning unfair ideas.
It's like if a robot only learned from people who thought some jobs pay less for women.
The best thing to do is be careful when using these robots and not believe everything they say.
A new study reveals that AI chatbots like ChatGPT exhibit gender and racial bias in salary negotiation advice.
Researchers analyzed LLMs like ChatGPT, Claude 3.5 Haiku, and GPT-4o mini and found biased recommendations.
The chatbots often suggested lower starting salaries for women compared to men in similar situations.
Bias was observed even when women possessed qualifications comparable to men.
Users are cautioned against solely relying on these models for information and advice.
- Who
- Researchers analyzed language models including Claude 3.5 Haiku, ChatGPT, and GPT-4o mini.
- What
- A study reveals AI models like ChatGPT exhibit gender and racial bias in salary negotiation advice.
- When
- Information about the study is recent.
- Where
- The study was conducted at Cornell University and the Technical University of Applied Sciences, Würzburg-Schweinfurt.
- Why
- To assess potential biases in AI models related to salary negotiation.
Study Findings
N/A
Bias in AI
Study Findings
AI models like ChatGPT provide biased salary advice.
N/A
The article doesn't present an opposing view, but this study identifies bias in AI models.
- Study Title
- Surface Fairness, Deep Bias: A Comparative Study of Bias in Language Models
- Key Finding
- Chatbots recommended lower salaries for women than men.
- Models Analyzed
- ChatGPT, Claude 3.5 Haiku, GPT-4o mini
- Researcher
- Iven P. Yamshchikov