Existential Risk Persuasion Tournament (XPT) Results
webA key empirical reference for quantitative existential risk estimates; frequently cited in AI safety discourse to contextualize AI risk relative to other global catastrophic threats.
Metadata
Importance: 72/100organizational reportdataset
Summary
The Existential Risk Persuasion Tournament (XPT) aggregated probabilistic forecasts from 169 participants—including domain experts, forecasting specialists, and superforecasters—on humanity's extinction risks by 2100. The tournament examined threats including AI, nuclear war, engineered pandemics, and other catastrophic risks, using structured deliberation and persuasion rounds to update estimates. It provides one of the most systematic crowd-sourced quantitative assessments of existential risk probabilities available.
Key Points
- •169 participants produced probabilistic forecasts on human extinction and civilization collapse risks by 2100 across multiple threat categories.
- •AI was identified as a major concern, with participants assigning notable probability to AI-related existential catastrophe.
- •The tournament used persuasion rounds where participants could update forecasts after reviewing arguments, testing whether expert deliberation shifts risk estimates.
- •Results offer rare quantitative benchmarks for comparing AI risk against other existential threats like nuclear war and engineered pandemics.
- •Findings are relevant to prioritization decisions in AI safety and global catastrophic risk funding and policy.
Review
The XPT represents an innovative approach to understanding complex existential risks by bringing together accurate forecasters and domain experts in a structured, collaborative prediction environment. By incentivizing participants to discuss, explain, and update their forecasts, the tournament aimed to generate high-quality insights into potential catastrophic scenarios facing humanity in the next century. The methodology's key strength lies in its interactive format, which allows participants to engage directly with different perspectives and potentially refine their predictions through structured dialogue. Of particular interest are the observed differences between superforecasters and expert perspectives, especially regarding the likelihood of catastrophic outcomes. The researchers noted intriguing discrepancies, such as why superforecasters seemed less concerned about extreme risks despite agreeing on many fundamental points. This approach provides a novel framework for exploring how expertise, forecasting skill, and interdisciplinary knowledge interact when assessing long-term global risks.
Cited by 3 pages
| Page | Type | Quality |
|---|---|---|
| Forecasting Research Institute (FRI) | Organization | 55.0 |
| XPT (Existential Risk Persuasion Tournament) | Project | 54.0 |
| Instrumental Convergence | Risk | 64.0 |
1 FactBase fact citing this source
| Entity | Property | Value | As Of |
|---|---|---|---|
| XPT (Existential Risk Persuasion Tournament) | Founded Date | 2022 | — |
Cached Content Preview
HTTP 200Fetched May 17, 20263 KB
Working paper
# Forecasting Existential Risks: Evidence from a Long-Run Forecasting Tournament
Forecasting Existential Risks: Evidence from a Long-Run Forecasting Tournament
What are the gravest risks facing humanity over the next century and beyond? In 2022, the Existential Risk Persuasion Tournament (XPT) brought together 80 experts and 89 superforecasters to predict the risks to humanity posed by artificial intelligence, nuclear war, pandemics, and other causes.
Ezra Karger†,\*, Josh Rosenberg\*, Zachary Jacobs\*, Molly Hickman\*, Rose Hadshar\*, Kayla Gamin\*, Taylor Smith\*, Bridget Williams\*, Tegan McCaslin\*, Stephen Thomas\*, Philip E. Tetlock‡,\* , Tap for info
† Federal Reserve Bank of Chicago
\\* Forecasting Research Institute
‡ Wharton School of the University of Pennsylvania
Published: Jul 10, 2023
Revised: Aug 8, 2023
Ezra Karger†,\*, Josh Rosenberg\*, Zachary Jacobs\*, Molly Hickman\*, Rose Hadshar\*, Kayla Gamin\*, Taylor Smith\*, Bridget Williams\*, Tegan McCaslin\*, Stephen Thomas\*, Philip E. Tetlock‡,\*
## Abstract
The Existential Risk Persuasion Tournament (XPT) aimed to produce high-quality forecasts of the risks facing humanity over the next century by incentivizing thoughtful forecasts, explanations, persuasion, and updating from 169 forecasters over a multi-stage tournament. In this first iteration of the XPT, we discover points where historically accurate forecasters on short-run questions (superforecasters) and domain experts agree and disagree in their probability estimates of short-, medium-, and long-run threats to humanity from artificial intelligence, nuclear war, biological pathogens, and other causes. We document large-scale disagreement and minimal convergence of beliefs over the course of the XPT, with the largest disagreement about risks from artificial intelligence. The most pressing practical question for future work is: why were superforecasters so unmoved by experts’ much higher estimates of AI extinction risk, and why were experts so unmoved by the superforecasters’ lower estimates? The most puzzling scientific question is: why did rational forecasters, incentivized by the XPT to persuade each other, not converge after months of debate and the exchange of millions of words and thousands of forecasts?
[Read the working paper](https://forecastingresearch.org/pdf/existential-risk-persuasion-tournament.pdf)
Acknowledgments
This research would not have been possible without the generous support of the Musk Foundation and the Long-Term Future Fund. We thank Walter Frick, Michael Page, Terry Murray, David Budescu, Barb Mellers, Elie Hassenfeld, Philipp Schoenegger, and Pavel Atanasov for their thoughtful feedback and comments. We are also grateful for the many people who reviewed earlier drafts of this report. We greatly appreciate the assistance of Amory Bennett, Kaitlyn Coffee, Adam Kuzee, Avital Morris, Fiona Pollack, Coralie Consigny, Arunim Agarwal, and Rumtin Sepasspour throughout the proje
... (truncated, 3 KB total)Resource ID:
5c91c25b0c337e1b | Stable ID: sid_Uwu1U4KZJn