Cool -- I will look at @edmundhandby.bsky.social and @nahshonp.bsky.social 's work. Send papers and ideas, please!

Thanks! Yes, it's like crtl-F, except it also tells you what words are frequent in the first place. Still. it's very crude. I don't think we learn much here -- but I wonder whether things get more interesting when one thinks about word connections.

(3) This is really just using AI for vibe coding for a quick-and-dirty analysis. It does not draw on other LLM capabilities. Everything here could be achieved with conventional coding or using specialist software (4) This benefits from others providing data. Thanks, Project Gutenberg! (5) Leviathan!

(1) This is not at all validated. It's really just playing around. Pinch of salt, etc. (2) It stems from me thinking about the uses of AI for political theory. To my mind, this is not a serious use -- but can LLMs be used for something serious in our discipline? Any thoughts welcome.

The good news is: if they do (despite their efforts), the cognitive gap between the groups shrinks. But what is going on? In-group solidarity, cognitive dissonance avoidance, self - deception? Sobering but fascinating work.

In this remarkable study, Shayo and co-authors show that subjects from israel pay to avoid reading about Palestinian victims while subjects from Jordan pay to avoid reading about Israeli victims. This is bad news.