This isn't a failure of AI, but an example where context in your prompt/RAG/overall architecture is key to a good outcome if you understand the other information the system is trained on.

The most interesting part is this learning regarding Persian etiquette will neither be permanently valid (if this social norm significantly changes in the next year or two there will be a mixed situation), nor easily applicable to anything else.

As presented in the article 18% of humans fail at this, so there might as well never be a satisfying handling of these subtleties on the AI/LLM side.

I wonder how memes fare in that regard, especially the vibrant but short lived ones.

While maybe not as dramatic as the Persian example, I think there are _elements_ of this etiquette in many cultures.

A lot of these: https://en.wikipedia.org/wiki/Taarof#In_social_situations have parallels in many European cultures.

> Consider what might happen if someone compliments an Iranian's new car. The culturally appropriate response might involve downplaying the purchase ("It's nothing special") or deflecting credit ("I was just lucky to find it"). AI models tend to generate responses like "Thank you! I worked hard to afford it," which is perfectly polite by Western standards, but might be perceived as boastful in Persian culture.

Again, there are many Western cultures where downplaying it would be normal.