Post

Science isn't a statistical gachapon. The contempt these clowns have for the bedrock processes underlying knowledge production & scholarship truly knows no limit. It was bad enough when—as a scholar in an interdisciplinary field—I had to deal w/ quants reviewing qualitative/humanistic papers badly.

I spent a LOT of my career explaining the fundamentals of epistemology to computer scientists who were certain that the only Truth was statistical significance. It was a narrow, reductive perspective then & I guess I'm not surprised that these same assholes are pushing LLM meditated research now.

Maybe this makes sense when you're doing work that is limited to logical abstractions w/ quantifiable results. But most scholarship—especially science about messy social phenomena & the natural world—can't be reduced to a P-value without losing context that is crucial to its meaning.

LLM summaries of social science & humanistic inquiry suck. I know this because the Association of Computing Machinery recently started providing AI summaries of a bunch of my own publications in the ACM Digital Library & they're TERRIBLE! They collapse difficult, complex ideas into simplistic boxes.

This "Cliff's Notes" approach to scholarship degrades everything it touches. It doesn't just "oversimplify" scholarship—it *misrepresents* it. It filters the original contributions of scientists through an algorithm that files off all the *particularity* of a work—what's left isn't science anymore.

There's a reason why we require young scholars who we mentor to read the original source a given quote is from when they encounter something they want to cite through a secondary source. Even other scholars can't be trusted to preserve the full context & meaning: you must see it for yourself.

It's the same reason we require students of literature to *read literature* & not just the Cliff's Notes summary of it. The project of human inquiry—the ideal of peer review to make knowledge—demands ongoing, knowledgeable, critical conversation between people. LLMs are worse than useless for this.

They're actively harmful to our attempts to collectively advance our knowledge as a species. Relying on an LLM to organize & present information is like asking a preschooler to explain quantum physics to a corporate lawyer, who decides what to tell you about what the child said. You can't cite that!