Forecasting research results

Abstract

The paper proposes the development of a platform to gather predictions from researchers, policymakers, and the public about the effects of social science interventions. It argues that gathering these ex-ante forecasts can be valuable in a number of ways. Firstly, it can provide additional information to supplement the limited data available on social interventions. Secondly, it can help researchers design better studies by identifying areas where their intuitions are particularly inaccurate. Thirdly, it can mitigate publication bias by providing evidence of the expected effect of an intervention, making null results more informative. The authors highlight the need for coordination in this area, as many researchers are independently seeking ex-ante forecasts from the same small pool of experts. They discuss the potential of the platform to track individual forecasters, identify areas where intuition is particularly accurate or inaccurate, and improve forecasting accuracy by de-biasing forecasts. The authors detail their current progress, including the beta version of the platform and plans for future development. – AI-generated abstract