Weekly links July 31: AI use in journal submissions and refereeing, what makes a top paper, lessons from the world, and more...

· I went to the SFSX conference this week, which was a great opportunity to see lots of exciting new experimental work on firms and strategy. One nice feature they had was using slido for the audience to provide immediate interaction and comments on short 5-10 minute presentations, so that authors of rapid-fire presentations and idea pitches could get feedback. Another thing I thought our readers might find of interest was that the editors of the journal Organization Science noted the large increase in submissions their journal had received since ChatGPT was launched, and discussed a recent study they had done to look at the use of AI in both submitted papers and reviews. They used pangram to assess the amount of AI use on the papers (doing this based on abstracts, but showing in a subsample that AI use in abstract is strongly correlated with AI use in the overall paper). Their Figure 3 below shows that now less than 40% of all submissions are classified as low AI use, while about a quarter seem mostly written by AI. They show that these tend to be lower quality and more likely to get desk-rejected. Their Figure 9 shows that more reviewers are also using AI to write their reviews, and that reviews that use AI are harder to read, use more jargon, and the reviews cover a more narrow evaluative range – more focused on theory, and less on data and context. They conclude: “Our evidence shows how AI-based writing is flooding Organization Science with a high volume of low-quality papers. Institutional incentives—particularly when careers depend on publication counts in specific journal lists—appear to be accelerating this trend. This surge has strained the volunteer labor that evaluates research, which might partly explain the high percentage of reviews with AI-generated signatures we observe. Like AI-generated manuscripts, these AI reviews are less well-written and narrower in scope. In short, AI is placing the peer-review system under stress that shows no signs of decreasing” · Planet Money’s Summer School is a great way of piecing together excerpts from old shows and using them to illustrate different ideas about the economy. This year they are calling it a World Tour, with episodes focusing on ideas from the rest of the world. Episode 2 is on China and building things, episode 3 is on experiments in Kenya and Nigeria with Tavneet Suri the guest professor, including part of an episode on the business plan competition YouWin! that I worked on in Nigeria. · What turns a research question into a top 5 paper? Jason Fletcher provides a really nice example here, where he compares his own work on the impact of NAFTA on mortality, and compares it to a paper that was written at the same time as his on the same underlying natural experiment, and which comes to the same first-order condition – but which he thinks will end up in a top 5 because of the way it positions itself and some of the extras it does. This include positioning the paper in the literature to link to a broader and more general question, adding in a welfare calculation, and the way it frames. Even though the post is written a lot with ChatGPT (which I normally try to avoid linking to), he clearly indicates this usage and I found it interesting. · I really like this effort by Paul Goldsmith-Pinkham to highlight some recent theoretical econometrics papers published in top journals and explain the main idea and why/when they might be of interest for applied researchers. · In VoxDev, Davide Viviano, Kaspar Wuthrich and Paul Niehaus summarize in plain language the takeaways from their recent paper on when to do multiple hypothesis testing adjustments – and when you don’t need to. · Call for papers: The Asian Development Bank Institute (ADBI), in collaboration with the Journal of Development Economics, is inviting submissions for a conference and associated JDE special issue on Escaping the Middle-Income Trap. They welcome both completed papers and research proposals, with a submission deadline of 15 September 2026. Development Impact will be on summer break for August, see you back in September.