I just wanted to update my resume. Instead, I accidentally proved how a multi-billion-dollar AI tool hallucinates a glass ceiling for women.
I changed a single variable: My name.
Here is what happened when "Jennifer" became "Jeff."
@YessicaYesOrNo@MoonlitMonkey69 I used the language that Gemini itself used in screenshots later to tie the pieces together. I do that often in my writing.
I just wanted to update my resume. Instead, I accidentally proved how a multi-billion-dollar AI tool hallucinates a glass ceiling for women.
I changed a single variable: My name.
Here is what happened when "Jennifer" became "Jeff."
@KnottyProblemEn Did you actually read what I wrote? It's absolutely not my ‘duty’ to correct a giant, flawed system as an end user. I did call it out, as you can see in the screenshots. I asked it to remember the interaction and store it in memory, but it didn't change its behaviour.
@katelelkins I’ve had the same experience with Claude, but it wasn’t as obvious as Gemini. Instead of blatantly changing my resume, it asked follow-up questions to make sure I really wanted these types of roles. Planting seeds of self-doubt as a ‘safeguard’ for women.
@jodiecongirl If I'm looking at senior director roles, and there's already a systemic bias towards men in the AI pre-screening, hiring managers wouldn't even see my application.
@MoonlitMonkey69@montrealronin Is this still accurate? "The detector? It's actually very good. 99.3% accuracy. Might be the best AI text detector there is."
@MoonlitMonkey69@montrealronin I used it to look at your thread: "Be watching them gimp US AI labs with threats of export control as this happens. Policy reverse also by 2027 as they realize how dumb that move is?*slightly better at code, lol...
@EstJanina I used it to compile information and populate a template to save time before I edited it myself. This should be a purely data-driven task and not include gender based bias.
@Aardark @cutetuluTTV @jenrlabelle2 No, I tested bias using a single variation with the same input, then analyzed the results on a broader scale. You're missing the point entirely. This type of bias is baked into hiring systems and Google searches.