Within Survey Estimates
How Hybrid Forecasting Shapes AI Doom Predictions
Structured forecasting events where experts discuss, debate, and revise their estimates of AI existential risk.
On this page
- Design of hybrid persuasion forecasting events
- Expert dialogue and probability adjustments
- Insights on confidence and consensus building
Page outline Jump by section
Introduction
Hybrid forecasting tournaments are structured events that blend formal probability elicitation with deliberative discussion to help experts refine their subjective estimates of complex, uncertain outcomes — including the probability of existential catastrophe from artificial intelligence (“AI doom”). In the context of long-tailed risks where historical data are scarce and reasoning is deeply uncertain, these tournaments are not just quizzes: they are carefully designed social‑cognitive experiments that reveal how experts think, why they disagree, and whether structured dialogue moves beliefs in practice. The most notable example so far — the Existential Risk Persuasion Tournament (XPT) — brings together specialist subject‑matter experts and forecasting generalists (“superforecasters”) to generate, discuss, and revise probability forecasts of catastrophic and existential risks, with AI risk as a major focal point. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
What Hybrid Forecasting Tournaments Are
Hybrid forecasting tournaments combine two elements that traditional expert surveys or markets typically separate:
- Quantitative probability forecasts: Participants assign numerical probabilities to clearly defined, verifiable outcomes — for example, whether a given risk threshold is reached by a target year.
- Structured deliberation and persuasion: Rather than making predictions in isolation, participants have rounds of argument exchange, reasoned debate, and explanation, intended to surface reasoning and encourage updating. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
This hybrid design builds on forecasting tournament methodology used in geopolitical and economic forecasting (where superforecasters have outperformed traditional analysts) but explicitly integrates adversarial collaboration — a form of structured debate aimed at reducing bias by having conflicting sides articulate and engage with each other’s reasoning. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
The Existential Risk Persuasion Tournament (XPT): A Case Study
Design of the Tournament
The Existential Risk Persuasion Tournament (XPT) ran from June to October 2022 and stands as the premier example of a hybrid forecasting contest aimed at existential threats including AI risk.
- Participants: 169 forecasters split roughly into two groups — 89 superforecasters with proven track records in probabilistic prediction, and 80 subject‑matter experts drawn from domains connected with long‑term risk (including AI safety, climate science, biosecurity and more). [Forecasting Research Institute]forecastingresearch.orgForecasting Research InstituteForecasting Existential Risks: Evidence from a Long-Run Forecasting Tournament – Forecasting Research Insti…
- Structure: Over four months, participants made initial anonymous forecasts; shared articulated reasons for their judgments; and engaged in multiple stages of debate and argument exchange aimed at challenging and refining those beliefs. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
- Questions: The tournament posed a mix of short‑term verifiable questions (about events between 2025–2030) and long‑term existential outcomes (such as probabilities of catastrophic loss or human extinction by 2100). [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
The key innovation was to blend the scoring accountability of forecasting tournaments with deliberative features borrowed from structured expert elicitation, seeking not just point estimates but insight into how reasoning evolves across dialogue. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
Expert Dialogue and Probability Adjustments
One of the most striking findings from the XPT was limited convergence of beliefs despite structured debate. Although superforecasters and domain specialists shared arguments and engaged with each other’s reasoning, there was little material movement in their respective forecasts across tournament stages. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
Two patterns emerged:
- Persistent divide on AI risk: Specialists consistently assigned higher probabilities to catastrophic and existential outcomes — particularly for long‑term AI risk — than superforecasters. The difference was larger for distant horizons than for near‑term questions. [IDEAS]ideas.repec.orgSource details in endnotes.
- Entrenched priors: Despite four months of argument exchange and incentives to influence peers, both groups tended to stick with their initial probability distributions rather than updating toward the other side’s view. This suggests strong cognitive anchoring and confirmation biases even among trained forecasters. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
These results provide a concrete lens on expert disagreement: it’s not just about who knows more, but about how different groups approach uncertainty and interpret sparse evidence. [Polycrisis]polycrisis.orgPolycrisisSubjective-Probability Forecasts of Existential Risk: Initial Results from a Hybrid Persuasion-Forecasting Tournament - Polycri…
Insights on Confidence, Consensus and Limits
Confidence and Calibration
One of the aims of hybrid forecasting tournaments is to shed light not only on point forecasts but on confidence calibration — how well probabilistic judgments align with actual outcomes where they resolve. Early follow‑ups to the XPT’s data (such as near‑term accuracy evaluations) find that superforecasters and domain experts both tend to outperform naive baselines in short‑range questions, but neither group shows a clear edge in long‑range forecasting accuracy. [Forecasting Research Institute]forecastingresearch.orgForecasting Research InstituteForecasting Existential Risks: Evidence from a Long-Run Forecasting Tournament – Forecasting Research Insti…
This points to a practical implication: while probability forecasts can be calibrated and compared on measurable questions, long‑run existential risk forecasting remains deeply uncertain even with structured methods. It also highlights that calibration on short horizons does not reliably translate into confidence about far‑future events. [Forecasting Research Institute]forecastingresearch.orgForecasting Research InstituteForecasting Existential Risks: Evidence from a Long-Run Forecasting Tournament – Forecasting Research Insti…
Consensus Building and Its Limits
The promise of hybrid tournaments is that structured discussion might reduce disagreement and improve collective beliefs. The XPT’s core finding — that sustained debate produced only modest within‑group shifts and limited cross‑group convergence — signals profound challenges:
- Cognitive biases remain: Even trained forecasters are susceptible to anchoring and confirmation bias, especially when stakes are high and evidence is thin. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
- Different epistemic approaches: Superforecasters often prioritise predictive calibration and base‑rate reasoning, while specialists may prioritise domain‑specific mechanisms and causal narratives. These frameworks do not always align, making consensus elusive. [Polycrisis]polycrisis.orgPolycrisisSubjective-Probability Forecasts of Existential Risk: Initial Results from a Hybrid Persuasion-Forecasting Tournament - Polycri…
This pattern matters for AI doom probability debates because it suggests that structured tournaments may reveal deeper structural divides in how experts think about existential risks, rather than magically resolving them. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
Why Hybrid Tournaments Matter for AI Doom Forecasting
Hybrid forecasting tournaments, like the XPT, contribute to the broader landscape of expert opinion in several ways:
- They make explicit the range of plausible probability estimates and why experts anchor them differently. [Polycrisis]polycrisis.orgPolycrisisSubjective-Probability Forecasts of Existential Risk: Initial Results from a Hybrid Persuasion-Forecasting Tournament - Polycri…
- They clarify where disagreement lies — for example, between empirical calibration and domain‑specific risk modelling in AI forecasts. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
- They provide a structured record of argumentation, useful for policymakers seeking transparent rationales behind divergent beliefs. [Polycrisis]polycrisis.orgPolycrisisSubjective-Probability Forecasts of Existential Risk: Initial Results from a Hybrid Persuasion-Forecasting Tournament - Polycri…
This methodological innovation complements traditional surveys by adding depth to how judgments are formed and where uncertainty stems from, rather than simply cataloguing point estimates of p(doom). [IDEAS]ideas.repec.orgSource details in endnotes.
Interpretation and Ongoing Challenges
While hybrid tournaments represent a promising frontier in expert elicitation, they also highlight the limits of forecasting long‑run, unprecedented events like existential catastrophe from AI:
- There is no substitute for evidence: When the underlying phenomena lack historical precedents, even structured debate struggles to generate convergence. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
- Design matters: Features like anonymity, incentives, question framing, and the mix of participants influence outcomes. Iterative improvements in design could yield richer insights in future tournaments. [Polycrisis]polycrisis.orgPolycrisisSubjective-Probability Forecasts of Existential Risk: Initial Results from a Hybrid Persuasion-Forecasting Tournament - Polycri…
- Interpretation requires care: Probability estimates from forecasting tournaments should be understood as subjective expressions of belief under uncertainty — useful for mapping expert views but not definitive predictions. [IDEAS]ideas.repec.orgSource details in endnotes.
Summary
Hybrid forecasting tournaments such as the Existential Risk Persuasion Tournament offer a novel, structured way to probe expert beliefs about AI doom probabilities. By combining numerical forecasting with sustained dialogue, they illuminate not only what experts estimate but why they differ and how (or whether) those estimates shift through argumentation. Early results show persistent disagreement, especially on long‑term AI risk, and limited belief convergence, underscoring both the potential and the profound challenges of using forecasting methods to refine assessments of existential risk. [ScienceDirect]sciencedirect.comScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie…
Amazon book picks
Further Reading
Books and field guides related to How Hybrid Forecasting Shapes AI Doom Predictions. Use these as the next step if you want deeper reading beyond the article.
The Precipice
Applies forecasting and uncertainty reasoning to global catastrophic risks.
Expert Political Judgment
Provides the research background behind expert prediction studies.
eBay marketplace picks
Marketplace Samples
Example marketplace items related to this page. Use the search link to explore similar finds on eBay.
Endnotes
-
Source: sciencedirect.com
Link: https://www.sciencedirect.com/science/article/pii/S0169207024001250Source snippet
ScienceDirectSubjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournament - Scie...
-
Source: ideas.repec.org
Link: https://ideas.repec.org/a/eee/intfor/v41y2025i2p499-516.html -
Source: polycrisis.org
Link: https://polycrisis.org/resource/subjective-probability-forecasts-of-existential-risk-initial-results-from-a-hybrid-persuasion-forecasting-tournament/Source snippet
PolycrisisSubjective-Probability Forecasts of Existential Risk: Initial Results from a Hybrid Persuasion-Forecasting Tournament - Polycri...
-
Source: youtube.com
Title: What superforecasters and experts think about existential risks | Ezra Karger
Link: https://www.youtube.com/watch?v=1WQzQ5hhz7kSource snippet
Existential Risk Persuasion Tournament...
-
Source: youtube.com
Title: Existential Risk Persuasion Tournament
Link: https://www.youtube.com/watch?v=_ApYZKYr4O4Source snippet
Who Was Liron Shapira BEFORE Doom Debates? - Interview on Theo Jaffee's Podcast from December 2023...
Published: December 2023
-
Source: forecastingresearch.org
Link: https://forecastingresearch.org/xptSource snippet
Forecasting Research InstituteForecasting Existential Risks: Evidence from a Long-Run Forecasting Tournament – Forecasting Research Insti...
-
Source: forecastingresearch.org
Link: https://forecastingresearch.org/research/near-term-xpt-accuracy -
Source: forecastingresearch.org
Link: https://forecastingresearch.org/near-term-xpt-accuracySource snippet
Assessing Near-Term Accuracy in the Existential Risk Persuasion Tournament — Forecasting Research InstituteASSESSING NEAR-TERM ACCURACY I...
Additional References
-
Source: wikidata.org
Link: https://www.wikidata.org/wiki/Q135880243Source snippet
WikidataSUBJECTIVE-PROBABILITY FORECASTS OF EXISTENTIAL RISK: INITIAL RESULTS FROM A HYBRID PERSUASION-FORECASTING TOURNAMENT (Q135880243...
-
Source: longtermwiki.com
Link: https://www.longtermwiki.com/projects/xptSource snippet
XPT (Existential Risk Persuasion Tournament) | Projects | Longterm WikiProjects/XPT (Existential Risk Persuasion Tournament) XPT (EXISTEN...
-
Source: stafforini.com
Link: https://stafforini.com/works/karger-2023-forecasting-existential-risks/ -
Source: jglobal.jst.go.jp
Link: https://jglobal.jst.go.jp/en/detail?JGLOBAL_ID=202502244992024278 -
Source: longtermwiki.com
Title: XP T (Existential Risk Persuasion Tournament) | Longterm Wiki
Link: https://www.longtermwiki.com/wiki/xptSource snippet
XPT (Existential Risk Persuasion Tournament) | Longterm WikiMarch 4, 2026 — XPT (EXISTENTIAL RISK PERSUASION TOURNAMENT) Project XPT (EXI...
Published: March 4, 2026
-
Source: longtermwiki.com
Title: XP T (Existential Risk Persuasion Tournament) | Longterm Wiki
Link: https://www.longtermwiki.com/wiki/E379Source snippet
XPT (Existential Risk Persuasion Tournament) | Longterm WikiJanuary 29, 2026 — WikiResponsesXPT (Existential Risk Persuasion Tournament)...
Published: January 29, 2026
-
Source: econpapers.repec.org
Title: RePEc:eee:intfor:v:41:y:2025:i:2:p:499 516
Link: https://econpapers.repec.org/RePEc%3Aeee%3Aintfor%3Av%3A41%3Ay%3A2025%3Ai%3A2%3Ap%3A499-516Source snippet
repec.orgEconPapers: Subjective-probability forecasts of existential risk: Initial results from a hybrid persuasion-forecasting tournamen...
-
Source: unjournal.pubpub.org
Link: https://unjournal.pubpub.org/pub/e2forecastingexistentialriskSource snippet
2 of "Forecasting Existential Risks: Evidence from a Long Run Forecasting Tournament", Applied Stream · The UnjournalAugust 23, 2024 — Pu...
Published: August 23, 2024
-
Source: youtube.com
Title: Who Was Liron Shapira BEFORE Doom Debates?
Link: https://www.youtube.com/watch?v=iRpRP0GIeLQSource snippet
Should you make and use existential risk estimates? How? | Michael Aird...
Published: December 2023
-
Source: youtube.com
Title: Should you make and use existential risk estimates? How? | Michael Aird
Link: https://www.youtube.com/watch?v=GYvKcZ-ZWXYSource snippet
Is AI Doom Going Out of Style?...
Topic Tree