Forecasting is the art and science of predicting the future. Forecasting can help us get a better sense of what will happen in the future, so that we can prepare appropriately. It is a significant aspect of EA strategy around pandemic preparedness (see biosecurity), AI alignment, animal product alternatives, and many other topics.
Forecasting is a feedback loop. Forecasters make predictions and then when those predictions resolve forecasters get scores. This allows the forecasters to improve and allows everyone else to know who the best forecasters are. In the future, we can weigh the forecasts of the better forecasters more heavily.
Tutorials
Alex Lawsen’s video series
Metaculus Tutorials
Further reading
Aird, Michael (2020) Failures in technology forecasting? A reply to Ord and Yudkowsky, LessWrong, May 8.
Dart Throwing Spider Monkey (2020) Intro to forecasting 01 - What is it and why should I care?, YouTube, October 26.
Kokotajlo, Daniel (2019) Evidence on good forecasting practices from the Good Judgment Project: an accompanying blog post, AI Impacts.
An excellent summary of the evidence on good forecasting practices.
Lewis, Gregory (2020) Challenges in evaluating forecaster performance, Effective Altruism Forum, September 8.
Muehlhauser, Luke (2016) Evaluation of some technology forecasts from “The Year 2000”, Open Philanthropy, September.
Muehlhauser, Luke (2019) How feasible is long-range forecasting?, Open Philanthropy, October 10.
Muehlhauser, Luke (2021) Superforecasting in a nutshell, Luke Muehlhauser’s Website, February 22.
Tetlock, Philip E. (2006) Expert Political Judgment: How Good Is It? How Can We Know?, Princeton: Princeton Univ. Press.
Tetlock, Philip E. & Dan Gardner (2015) Superforecasting: The Art and Science of Prediction, New York: Crown Publishers.
Vivalt, Evalt (2020) Announcing the launch of the Social Science Prediction Platform!, Eva Vivalt’s Blog, July 7.
Wiblin, Robert (2017) Prof Tetlock on predicting catastrophes, why keep your politics secret, and when experts know more than you, 80,000 Hours, November 20.
Wiblin, Robert & Keiran Harris (2019) Accurately predicting the future is central to absolutely everything. Professor Tetlock has spent 40 years studying how to do it better, 80,000 Hours, June 28.
External links
Forecasting Wiki. A wiki about forecasting and all things related to it.
Related entries
AI forecasting | credence | estimation of existential risk | Forecasting Newsletter | improving institutional decision-making | inside vs. outside view | long-range forecasting | Metaculus | model uncertainty | prediction markets | tabletop exercises | value of information
I think it’d be good for an editor to add a section on specification problems, drawing on this post, as Aaron suggests in a comment there.
But unfortunately:
I don’t have time to do this myself
It might be a bit weird to do this before the entry is expanded in other ways, since there’s also a lot else that could be useful to say here?
E.g., this article could add sections based on Daniel K’s summary of evidence from Good Judgement project, could add stuff about “CHAMPS KNOW”, could add some brief mention of long-range forecasting and link to the relevant entry, etc.