Another day, another excellent demonstration of why doing data analysis using #GenAI is not only unethical but also completely nonsensical: https://kucharski.substack.com/p/real-signals-or-artificial-stereotypes.
The author got Microsoft Copilot to analyse identical but randomly ordered datasets of textual responses assigned to different countries. Copilot produced a plausible-sounding summary of differences between the countries that are nothing less than textbook examples of country stereotypes.
Example curtsy of @BerLinguistin via LinkedIn.