{
  "id": 17097922,
  "title": "Can Artificial Intelligence Forecast Treatment Effects in Economics?",
  "url": "https://urgent.news/2026/10/29/can-artificial-intelligence-forecast-treatment-effects-in-economics",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-29T18:02:03.000Z",
  "source": {
    "name": "Marginal Revolution",
    "slug": "marginal-revolution",
    "url": "https://marginalrevolution.com/marginalrevolution/2026/10/can-artificial-intelligence-forecast-treatment-effects-in-economics.html"
  },
  "original_language": "en",
  "account": "Artificial intelligence (AI) models are being employed to predict the outcomes of research studies that have been pre-registered on platforms such as the American Economic Association's Research Triangle Registry (AEA RCT Registry) and the Open Science Framework. These models concentrate on studies whose results were initially made public in 2026 or later.\n\nThe AI models begin by extracting specific forecasting questions from the pre-registrations or pre-analysis plans. Although not all of these questions result in forecasted treatment effects, those that do are utilized to evaluate the accuracy of the forecasts. In total, forecasts were generated for 889 treatment effects from 39 papers with results disclosed in January 2026 or later, and 673 forecasts from 33 papers after screening.\n\nThese forecasts are initially provided in raw units, but they can be standardized, thereby allowing for comparisons among different studies. Currently, the median absolute forecast error stands at 0.054 standard deviations for the best-performing model, with an average absolute error of 0.093 standard deviations and a mean squared error of 0.0218. It is noteworthy that AI forecasts outperform benchmarks that predict no effect or simply forecast the mean effect of other papers' results within the same timeframe.\n\nThese findings indicate that there is significant potential for AI forecasts to play a valuable role in informing both policy and research decisions. As Eva Vivalt, a leading young economist, points out, much of economics (and science) is undergoing radical changes, and this AI forecasting method may be at the forefront of this transformation. The post highlights the importance of shifting focus towards projects with high uncertainty and significant upside potential.",
  "summary": "I use LLMs to forecast treatment effects of research studies pre-registered on the AEA RCT Registry and the Open Science Framework. I focus on studies whose results were first publicly released in 2026 and elicit forecasts from models whose training or knowledge cut-off dates were before the relevant set of papers was released. First, LLM […] The post Can Artificial Intelligence Forecast…",
  "key_points": [
    "AI models predict treatment effects in pre-registered studies.",
    "889 forecasts generated for 39 papers with 2026+ results.",
    "AI forecasts outperform benchmarks predicting no effect."
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}