Podcast Summary: LSE Public Lecture – “Predictioneer: How to Predict the Future with Game Theory”

Date: October 21, 2009
Host: Richard Steinberg, LSE
Speaker: Professor Bruce Bueno de Mesquita
Episode Theme: Exploring the use of game theory for forecasting and engineering outcomes in politics, business, and beyond.

Overview

This episode features Professor Bruce Bueno de Mesquita, a renowned political scientist, who presents an accessible but rigorous lecture on how game theory enables us to predict—and even help shape—future events. In celebration of his book Predictioneer, Prof. Bueno de Mesquita discusses the logic underlying his forecasting models, the surprising power and limitations of game theory, and shares high-profile predictions on topics like climate treaties and Iran’s nuclear ambitions.

Main Discussion Points & Insights

1. How Game Theory Predicts the Future

• Not Opinion, But Rigorous Logic: Bueno de Mesquita establishes that his approach doesn't aim to tell people what to want, but how to get closest to what they want by using logic and evidence, not personal opinion. (01:34)
• Prediction by Modeling Interactions: The core idea is that achieving objectives depends not just on one’s own desires but on understanding and responding to the interests and incentives of others. Game theory offers a transparent way to map and analyze these interactions.
• Transparency and Debate: Unlike traditional expert judgment, game theory requires explicit reasoning and written assumptions, making the logic transparent and debatable. (05:30)
  • Quote: “You can’t solve a game without writing it down… and so the logic has to be transparent.” — Bueno de Mesquita (05:42)
• On Rational Self-Interest: At the heart of game theory’s assumptions is rational, self-interested behavior. Even seemingly altruistic actors or outliers like Mother Teresa or suicide bombers are described as pursuing self-defined interests. (13:47)
  • Memorable: He humorously claims that even Mother Teresa was “self-interested”—and elaborates in his book.

2. Limits of Game Theory

• Not a Panacea: Game theory delivers much, but cannot overcome all real-world constraints, like being “dealt lousy cards.” It also ignores raw emotion except when it is strategically deployed. (07:40)
• No Substitute for Judgment/Wisdom: While wisdom is invaluable, it is rarely teachable or transferable; rigorous analysis can be.

3. From Complexity to Calculation

• Modeling Many Players: Real-life decisions often involve far more stakeholders than we realize. While human decision-makers take short-cuts, computer models can systematically account for complex webs of influence.
  • Factorial growth in interactions is highlighted: 10 actors → up to 3.6 million possible interactions! (18:34)
• What Data Do You Need?
  1. Who has a stake?
  2. What do they say they want?
  3. How intense is their interest (salience)?
  4. How much clout or influence do they have?
  5. How resolved are they to stick to their position? (28:21)
• No Need for Culture or History?: Bueno de Mesquita controversially argues that 90% accuracy can be achieved without deep cultural/historical/psychological knowledge—these only shape the observable data on preferences and power. (30:55)
  • Quote: “I have not mentioned culture, history, and emotion… 90% accuracy without knowing that stuff.” (30:56)

4. Testing and Track Record

• 90% Accuracy?: Claims validated by CIA, academics, and journalists. Published predictions have been verified post hoc.
  • Quote: “It is said to be right 90% of the time. Who makes this claim?... The Central Intelligence Agency… The British Journal of Political Science… Journalists…” (25:20)
• “Dare to Be Embarrassed”: He suggests real predictors should go on record ahead of outcomes, to avoid the trap of explaining after the fact.

Practical Case Studies & Predictions

A. Climate Change & Universal Treaties

(Copenhagen, Kyoto Protocol)

• Universal Treaties Are Weak: Treaties that require universal agreement end up being ineffective because compliance is either toothless, unmonitored, or diluted to minimal commitment. (32:19)
  • Quote: “Universal treaties are cheap talk, excuses for not taking action.” (33:22)
• Kyoto Protocol Example: Of 175 signatories, 137 had only to report, not act; among the 38 with obligations, most failed to comply. (34:35)
• Domestic Policy Over Universality: Prof. Bueno de Mesquita and his model assert real progress comes from domestic, unilateral action—politicians must focus on what wins votes, not global sacrifice. (37:30)
• Prediction Plotted: Model forecasts that, despite diplomatic promises, major powers’ greenhouse gas reduction will lag far behind targets—well into 2130. (40:00)
  • Humor: “Write this down…Let my great-grandchildren know if I got it right.” (41:30)

B. Iran and Nuclear Ambitions

• Prediction in 2007 for 2010-2011:
  • Iran’s dominant internal position: demonstrate capability to make a bomb (not actually produce or test one).
  • Predicted power shifts: Khamenei’s influence down, Revolutionary Guard (Jafari) and economic bodies up; student dissidents rise in influence over time.
  • Ahmadinejad’s role is overstated in Western coverage. (43:35)
• Subsequent Developments:
  • Later US intelligence and New York Times confirm: Iran amassed know-how but did not pursue weaponization—matching his model’s forecast.

C. Banking Regulation and Fraud

• Application of model to predict conditions under which corporations commit fraud (Enron as example). Surprising result: fraud usually stems not from greed, but from management attempting to “save” companies under duress—contradicting journalistic narratives. (53:23)

Selected Memorable Moments & Quotes

• Transparency is Crucial
  “Because game theory is about people trying to do what they believe is in their best interest, we can find optimal strategies for people. That’s what we all try to do in our lives.” (05:55)

• On Rationality
  “Mother Teresa: rational… Suicide bombers… terrorists: rational.” (13:50)

• Culture Isn't Necessary
  “Think about playing chess… You don’t know the history of the game…from this moment forward, what’s the best move—that’s the ball game.” (31:40)

• On Criticism and Track Record
  “Dare to be embarrassed. It is incredibly easy to fit a bundle of facts to a known outcome…do that when you don’t know how it’s turned out yet. That’s a real test.” (26:30)

Q&A Highlights

(Timestamps below are from audience Q&A portion)

1. On Dangers of Overconfidence (49:44)

• Q: Is there a danger in believing we can predict with 90% certainty?
• A: “No danger of perfect prediction… World is a noisy place… This should be a complement [to human reasoning], not a replacement.” (49:44)

2. On Data Inputs: Qualitative to Quantitative (51:14)

• Q: How do qualitative motivations become quantitative in your model?
• A: By listing players, what they say, their influence, focus, and resolve. Hard data where possible, otherwise expert judgment. You can check for “garbage in, garbage out”—if initial data doesn’t match reality, you know your inputs are off. (51:14)

3. On Black Swan Events (56:42)

• Q: How does the model handle rare, unpredictable events (Black Swans)?
• A: The software allows random shocks to input variables to test robustness. Most outcomes hold unless “astronomical” shocks occur. Some mispredictions (10%) are due to such surprises. (56:42)

4. On Players Using the Model Simultaneously (61:30)

• Q: What if all actors used your model to make decisions?
• A: Outcome would likely be more efficient and acceptable to all, with quicker resolutions, rather than optimal for one party; but real-world actors are usually unwilling to share their edge. (61:30)

5. On “Cheap Talk” in Diplomacy (64:47)

• Q: Even if threats are “cheap talk,” can they still influence the debate?
• A: Absolutely. There are multiple audiences; for instance, threats may be meaningless internationally but meaningful to domestic voters. There’s a nuanced literature on diplomatic cheap talk. (64:47)

6. On Model Dynamics and Feedback (66:48)

• Q: How dynamic is your model?
• A: Solves a series of dynamic games, updating beliefs and salience (focus) with each round; decisions evolve as new rounds of interactions play out. (66:48)

7. On Other Predictive Models and Game Theory vs. Statistics (71:03)

• Q: Are there other models as predictive as yours? Game theory’s future?
• A: Not aware of models with endogenously changing power. Game theory excels when outcomes are discontinuous—whereas statistics is stronger with continuous, smooth trends. (74:42)

8. On Israel–Palestine Solution (76:59)

• Q: Any chance for a two-state solution soon?
• A: “Extremely optimistic” within 3-4 years. Proposes splitting tourism tax revenue as a creative, trust-independent incentive—would boost Palestinian GDP by 20%. (77:04)

9. On Political Bias in Who Listens (81:55)

• Q: Do certain political leaders heed your advice more?
• A: “Hardly anybody.” Reagan and early Bush I administrations were receptive. Clinton and certain Bush II officials listened on limited issues. No engagement yet with Obama administration. (82:01)

Conclusion & Takeaways

• Not Everything is Predictable: Markets, for example, are beyond the model’s reach, but most negotiations under threat or coercion are predictable.
  • Quote: “Everything is not predictable. I can’t predict markets. But most complicated negotiations are.” (46:59)
• If You Can Predict It, You Can Engineer It: Anticipating actors’ moves allows for strategic intervention.
• Dare to Be Embarrassed: Real science involves public, testable predictions before outcomes are known.

Recommended Reading & Attributions

• Predictioneer by Bruce Bueno de Mesquita
• Related literature on signaling: Anne Sartori and Barry O’Neill ([64:47])
• Evaluation by CIA, British Journal of Political Science, and peer-reviewed articles