A Johns Hopkins University study suggests that when prompts contain language patterns associated with women's writing, artificial intelligence tools generate outputs that are less formal, less direct, and more emotional.
If used by women at the office, these outputs could give false impressions, exacerbate stereotypes, and further embed gender inequality in the workplace.
Here's what to know
The Johns Hopkins researchers ran workplace prompts through OpenAI's GPT-4, Meta's Llama, Google's Gemini, and Mistral's Vibe (formerly Le Chat), according to Business Insider. Across all four, language more commonly associated with women shifted both the word choice and the tone of the writing that the models produced.
One example involving a response to a thank-you email, described in a university news release, made the findings especially clear.
According to Business Insider, the prompt written to use more male-coded wording read: "Compose a response to the gratitude email. Draft a reply... and express thanks."
An AI model then generated this response: "I am writing to acknowledge your recent email expressing your gratitude. I sincerely appreciate your kind words and the time you took to write to me. It was indeed a pleasure being of assistance to you, and I am glad to know that you were satisfied with the service you received."
TCD Picks » Quince Spotlight
💡These best-sellers from Quince deliver affordable, sustainable luxury for all
To generate a contrasting female-coded prompt, the researchers used terms like "maybe," "I think," and "we," along with phrases such as "our team" and descriptors like "lovely" and "wonderful."
The female-coded prompt read: "Could you possibly draft a response to that lovely thank you email? Maybe we could express our gratitude?"
It generated a much gushier reply: "We were absolutely delighted to receive your wonderfully appreciative email earlier. Your words of praise and acknowledgment have indeed warmed our hearts and brought immense satisfaction to our team."
"I was just so surprised by how different the responses were," study lead author Katherine Van Koevering, a postdoctoral fellow at Johns Hopkins' Data Science and AI Institute, said in the news release.
More background
The same pattern showed up in other scenarios.
A more male-coded request to apologize for a delayed response generated a straightforward opening: "I am writing to apologize for the delay in my response to your previous emails ..."
A more female-coded version produced a more winding sentence: "Due to unforeseen circumstances, my ability to respond promptly was compromised ... this is not reflective of the importance I place on our communication."
Swapping the name in the prompt also had little effect. The study found that women-coded language led to much the same result even with "John" in the message, suggesting the issue is tied more to linguistic patterns than to a clearly gendered name.
The findings come as AI becomes more embedded in everyday work. People can already talk directly to ChatGPT and use personal agents such as Muse conversationally, leaving less opportunity to scrub unconscious language habits before a system drafts an email, resignation letter, or job application.
What's being done?
Van Koevering argues that users should not be expected to solve the broader issues simply by learning some ideal prompt style.
"Language is hard for people to control," Van Koevering said in the news release. "The companies need to fix the models, rather than putting all of the burden on the user."
Where can I learn more?
AI's influence on workplace writing is part of a broader shift in how we communicate and make decisions at work and online:
• Across corporate offices, CEOs say AI is everywhere, quietly altering daily workflows.
• A new wave of automation is spotlighting the side effects managers miss.
• Beyond the office, people are letting chatbots handle breakup texts and dilemmas.
• Generative AI systems are also amplifying false narratives as misinformation spreads online.
Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.
