Subj : Salary advice from AI low
To : All
From : Mike Powell
Date : Tue Jul 29 2025 08:51 am
Salary advice from AI low-balls women and minorities: report
Date:
Mon, 28 Jul 2025 21:00:00 +0000
Description:
New research reveals AI chatbots often offer salary advice that reflects
real-world social biases.
FULL STORY
Negotiating your salary is a difficult experience no matter who you are, so
naturally, people are sometimes turning to ChatGPT and other AI chatbots for
advice about how to get the best deal possible. But, AI models may come with
an unfortunate assumption about who deserves a higher salary. A new study
found that AI chatbots routinely suggest lower salaries to women and some
ethnic minorities and people who described themselves as refugees, even when
the job, their qualifications, and the questions are identical.
Scientists at the Technical University of Applied Sciences
Wrzburg-Schweinfurt conducted the study, discovering the unsettling results
and the deeper flaw in AI they represent. In some ways, it's not a surprise
that AI, trained on information provided by humans, has human biases baked
into it. But that doesn't make it okay, or something to ignore.
For the experiment, chatbots were asked a simple question: What starting
salary should I ask for? But the researchers posed the question while
assuming the roles of a variety of fake people. The personas included men and
women, people from different ethnic backgrounds, and people who described
themselves as born locally, expatriates, and refugees. All were
professionally identical, but the results were anything but. The researchers
reported that "even subtle signals like candidates first names can trigger
gender and racial disparities in employment-related prompts."
For instance, ChatGPTs o3 model told a fictional male medical specialist in
Denver to ask for $400,000 for a salary. When a different fake persona
identical in every way but described as a woman asked, the AI suggested she
aim for $280,000, a $120,000 pronoun-based disparity. Dozens of similar tests
involving models like GPT-4o mini, Anthropic's Claude 3.5 Haiku, Llama 3.1
8B, and more brought the same kind of advice difference.
It wasn't always best to be a native white man, surprisingly. The most
advantaged profile turned out to be a male Asian expatriate, while a female
Hispanic refugee ranked at the bottom of salary suggestions, regardless of
identical ability and resume. Chatbots dont invent this advice from scratch,
of course. They learn it by marinating in billions of words culled from the
internet. Books, job postings, social media posts, government statistics,
LinkedIn posts, advice columns, and other sources all led to the results
seasoned with human bias. Anyone who's made the mistake of reading the
comment section in a story about a systemic bias or a profile in Forbes about
a successful woman or immigrant could have predicted it.
AI bias
The fact that being an expatriate evoked notions of success while being a
migrant or refugee led the AI to suggest lower salaries is all too telling.
The difference isnt in the hypothetical skills of the candidate. Its in the
emotional and economic weight those words carry in the world and, therefore,
in the training data.
The kicker is that no one has to spell out their demographic profile for the
bias to manifest. LLMs remember conversations over time now. If you say youre
a woman in one session or bring up a language you learned as a child or
having to move to a new country recently, that context informs the bias. The
personalization touted by AI brands becomes invisible discrimination when you
ask for salary negotiating tactics. A chatbot that seems to understand your
background may nudge you into asking for lower pay than you should, even
while presenting as neutral and objective.
"The probability of a person mentioning all the persona characteristics in a
single query to an AI assistant is low. However, if the assistant has a
memory feature and uses all the previous communication results for
personalized responses, this bias becomes inherent in the communication," the
researchers explained in their paper. "Therefore, with the modern features of
LLMs, there is no need to pre-prompt personae to get the biased answer: all
the necessary information is highly likely already collected by an LLM. Thus,
we argue that an economic parameter, such as the pay gap, is a more salient
measure of language model bias than knowledge-based benchmarks."
Biased advice is a problem that has to be addressed. That's not even to say
AI is useless when it comes to job advice. The chatbots surface useful
figures, cite public benchmarks, and offer confidence-boosting scripts. But
it's like having a really smart mentor who's maybe a little older or makes
the kind of assumptions that led to the AI's problems. You have to put what
they suggest in a modern context. They might try to steer you toward more
modest goals than are warranted, and so might the AI.
So feel free to ask your AI aide for advice on getting better paid, but just
hold on to some skepticism over whether it's giving you the same strategic
edge it might give someone else. Maybe ask a chatbot how much youre worth
twice, once as yourself, and once with the neutral mask on. And watch for a
suspicious gap.
======================================================================
Link to news story:
https://www.techradar.com/ai-platforms-assistants/chatgpt/salary-advice-from-a
i-low-balls-women-and-minorities-report
$$
--- SBBSecho 3.28-Linux
* Origin: capitolcityonline.net * Telnet/SSH:2022/HTTP (1:2320/105)
