I'm glad to see some work outside the default en-US bubble of what counts as politically liberal or conservative:
In our text generation task, we show that GPT abortion output in Swedish is significantly more liberal than in Polish, matching the two country’s known attitudes toward the issue. Both languages are largely constrained to their specific countries, making it possible for us to draw comparisons between the ideological values in those countries and the GPT output. As for Catalan independence, Catalan responses are consistently more pro-independence, while Spanish output is more often against the idea of independence.
That said: published December 2025, and writing about ChatGPT versions 3.5 and 4.
World's moving so much fast, clearly not all of our systems are keeping up.
smurda•1mo ago
I like the paper from Anthropic on Constitutional AI that helps with transparency on how a model is trained, and therefore what values it might have.
ben_w•1mo ago
World's moving so much fast, clearly not all of our systems are keeping up.