New working paper: AI can predict economics treatment effects from pre-registrations

Economist Eva Vivalt released a new working paper on October 29 asking a core question: can AI predict the treatment effects of economics (field) experiments from their pre-registration documents alone, before the experiments are run? The paper's conclusion is yes—models can produce valuable effect predictions based solely on pre-registration content.

Confirmed

• The paper's author is Eva Vivalt; it is a working paper (not a formally published version) that circulated on social media around October 29, sparking heavy reposting and discussion.
• The test subject is LLM/AI, with pre-registration documents as the only input and predictions of experimental treatment effects as output; the conclusion is that prediction is possible.
• All reposts are secondhand accounts that provide no methodological details, data scale, or prediction accuracy figures.

Why it matters

• Reposter abhishekn called it the "craziest economics result" he has seen this year, saying it will push the field to rethink existing research workflows.
• Economist John Horton reposted that "this is a very important paper," noting major implications if the findings hold (the post did not elaborate on mechanisms).
• Ethan Mollick commented from a research-paradigm angle: scholars invest enormous effort in individual papers and thus favor low-risk research questions; this result suggests AI might forecast conclusions before experiments are run, disrupting how research topics are chosen and evaluated.
• paulnovosad noted in a reply that this echoes a view in economics—since LLMs can already handle desk research, the relative importance of randomized controlled trials (RCTs) will rise.

Not yet confirmed

• None of the reposts disclose the paper's specific experimental design, models used, or prediction accuracy; the judgments above rest on the paper's overall conclusion, and how well it actually holds up awaits the full paper and follow-up validation.

2026-10-29 ~ 2026-10-29 · 5 related posts

• New Working Paper: AI Can Forecast Treatment Effects in Economics From Pre-Registration Docs — birchlse · 2026-10-29
• New Working Paper Finds LLMs Can Forecast Economics RCT Treatment Effects — soumitrashukla9 · 2026-10-29
• Ethan Mollick: AI Can Forecast Economics Treatment Effects, Rethinking How Science Is Done — soumitrashukla9 · 2026-10-29

2 near-duplicate retellings: soumitrashukla9 · soumitrashukla9