Economist Novosad warns AI is degrading academic writing and young researchers' thinking

On October 15, development economist Paul Novosad (about 180K followers) posted a series of tweets systematically criticizing AI's erosion of academic writing and research, which spread widely. Anna Stansbury retweeted in agreement, arguing that AI performs poorly in both prioritization and adherence to academic writing norms when drafting and summarizing academic economics content.

Confirmed

Novosad believes the most damaging flaw in current AI-assisted academic writing is the inability to tell what matters from what doesn't. This failure of judgment is not only the most glaring tell of AI use, but also reveals the author's own poor judgment—and it is extremely hard to fix through editing alone, short of extensive rewriting.
He points to a classic scenario: starting from a weak idea, discussing it with people leads to better ideas, whereas AI will dress up a bad idea in impressive economics jargon and contract theory, making it harder to recognize as bad while looking sophisticated.
He criticizes how early-career researchers are being led astray: reading AI summaries instead of the original papers, talking to AI instead of working through ideas on paper, and ending up steered in wrong directions by AI's mediocre judgment.
His advice: AI may improve, but for now it is leading many young people to produce worse work and worse thinking. Junior researchers should read papers and write out their own ideas; AI can be used for literature search, but should be kept out of the thinking process.
He advocates a writing style where any university-educated reader can grasp the core idea and its value, singling out QJE (Quarterly Journal of Economics) as consistently excellent on this front.

Why it matters

This is not a generic "AI plagiarism" critique, but points to a deeper risk: what AI erodes is researchers' judgment and idea generation—the very core of academic work—and it harms early-career researchers, who lack the discernment to compensate, the most.
Coming from an influential scholar and echoed by fellow economists, the view reflects growing alarm within economics about the downsides of AI-assisted research, with lessons for researchers in other fields as well.

2026-10-15 ~ 2026-10-15 · 7 related posts

Primary sources

Economist flags AI's worst academic 'tell': bad judgment on what matters — paulnovosad · 2026-10-15 22:00 (UTC+08:00)
AI dresses up weak ideas in jargon, luring junior researchers into worse thinking — paulnovosad · 2026-10-15 22:00 (UTC+08:00)
Advice for junior researchers: read papers and write without AI in the loop — paulnovosad · 2026-10-15 22:00 (UTC+08:00)

[source] Economist flags AI's worst academic 'tell': bad judgment on what matters — paulnovosad · 2026-10-15
Economist on AI writing's worst tell: bad judgment about what matters — paulnovosad · 2026-10-15
Economist warns: AI summaries are quietly steering young researchers astray — paulnovosad · 2026-10-15
[source] AI dresses up weak ideas in jargon, luring junior researchers into worse thinking — paulnovosad · 2026-10-15
[source] Advice for junior researchers: read papers and write without AI in the loop — paulnovosad · 2026-10-15
Economist Paul Novosad: Use AI to Find Papers, Keep It Out of Your Thinking — paulnovosad · 2026-10-15
Economists: AI's worst academic-writing tell is poor judgment on what matters — soumitrashukla9 · 2026-10-15