The AI revolution has hit economists. I’ve seen offers of AI-generated seminar practice, allowing them to field hostile questions from a virtual “Dr Robust”. I’ve heard about a project to use AI to write 1,000 policy evaluations. And just this week, I read my favourite economics newsletter (Best of Econtwitter) scolding contributors for posting AI slop on social media: “Stop that!” So, are these new tools making economics better?
There are at least three different ways that AI could enhance the discipline. First, it could make economists enormously productive, as it becomes trivially easy to churn out existing methods of research. Second, it could explode the scope of economic inquiry. And third, it could help to weed out mistakes.
Taking each in turn, it’s clear that AI has huge potential to automate away tedious tasks such as data cleaning, grant writing and table formatting. (Having cleaned a fair amount of data in my time, good riddance.) Adopters are thrilled. Paul Novosad of Dartmouth College told me that he has roughly quintupled the time he could spend actually thinking about research questions. Elliott Ash of ETH Zürich university told me that the productivity gains were so exciting, it made him want to work more.