For anyone else also interested, I went and had a look at the links Dessalines kindly provided.

The source on the graphs says “Sources: Daniel Cox, Survey Center on American Life; Gallup Poll Social Series; FT analysis of General Social Surveys of Korea, Germany & US and the British Election Study. US data is respondent’s stated ideology. Other countries show support for liberal and conservative parties All figures are adjusted for time trend in the overall population.” Where FT is financial times.

It’s not clear how the words “liberal” and “conservative” were chosen, whether they’re intended to mean “socially progressive” and “socially traditional” or have other connotations bound with the political parties too, and whether the original data chose those descriptions or if they’re FT’s inference as being “close enough” for an American audience.

Unfortunately the FT data site is refusing to let me look at them without “legitimate interest” advertising cookies so I can’t tell you much more or if there’s any detail on methodology.

How convenient that a counterexample can’t be named

I feel like Luthor was a better counterexample for this before the model for his billionaire redesign was elected President of the USA.

Even so, Luthor hasn’t had quite the same volume of appearances as Iron Man, Batman, Captain America and the other rich superhero tropes.

People have grown up reading comic books and watching movies about generous billionaire superhero saviors. They want to believe that exists because it’s what they’ve been taught justice looks like.

You’re both adorable

References weren’t paywalled, so I assume this is the paper in question:

Hofmann, V., Kalluri, P.R., Jurafsky, D. et al. AI generates covertly racist decisions about people based on their dialect. Nature (2024).

Abstract

Hundreds of millions of people now interact with language models, with uses ranging from help with writing^1,2 to informing hiring decisions³. However, these language models are known to perpetuate systematic racial prejudices, making their judgements biased in problematic ways about groups such as African Americans^4,5,6,7. Although previous research has focused on overt racism in language models, social scientists have argued that racism with a more subtle character has developed over time, particularly in the United States after the civil rights movement^8,9. It is unknown whether this covert racism manifests in language models. Here, we demonstrate that language models embody covert racism in the form of dialect prejudice, exhibiting raciolinguistic stereotypes about speakers of African American English (AAE) that are more negative than any human stereotypes about African Americans ever experimentally recorded. By contrast, the language models’ overt stereotypes about African Americans are more positive. Dialect prejudice has the potential for harmful consequences: language models are more likely to suggest that speakers of AAE be assigned less-prestigious jobs, be convicted of crimes and be sentenced to death. Finally, we show that current practices of alleviating racial bias in language models, such as human preference alignment, exacerbate the discrepancy between covert and overt stereotypes, by superficially obscuring the racism that language models maintain on a deeper level. Our findings have far-reaching implications for the fair and safe use of language technology.