STAGE 2 · MODULE 2.4

Game Theory & Mechanism Design — Captivity & Escape

Game theory: the calculus of obedience, cooperation, and exit. Mechanism design: the inverse problem — install rules so self-interest produces the intended equilibrium.

0. Frame & Lineage

Game theory formalizes strategic interaction: when outcomes depend on what others do (and what they believe you’ll do). It tells you when obedience is rational, when cooperation is rational, and when “rationality” is being produced by an installed mechanism rather than chosen freely.

Mechanism design runs the arrow backward: choose rules (messages, constraints, enforcement) so that equilibrium behavior implements the intended outcome.

Reading key: treat every concept as “under engineered preferences, bounded rationality, incomplete information, and an adversarial rule-writer.”
Primary backbone (course) video course Nash · repeated · signaling · auctions
Ben Polak’s Open Yale course is the clean, broad scaffold for the full module.
Origin + culture (where this came from) war / control lineage von Neumann
Founding text + a Cold War–era documentary to see the cultural substrate that birthed “strategic man.”
Canon nodes (named architects) von Neumann · Nash · Selten · Harsanyi · Schelling · Shapley · Hurwicz · Maskin · Myerson · Axelrod · Binmore · Nisan · Roughgarden · Aumann · Hart · Kahneman/Tversky

Use these names as handles for the machinery; no reverence required.

1. What is a “game” — and what’s being smuggled in?

Formally: players, strategies, payoffs, information, timing, and a solution concept (Nash, correlated equilibrium, etc.).

Hidden assumptions (usually unspoken):
  • Preferences are exogenous (not engineered).
  • Rationality is common knowledge (not bounded/hijacked).
  • Rules are fixed (no live rule-writer patching mid-game).
Operational rule for this module: read each model as “a lens on incentives + information + enforcement,” then ask who controls those three levers.
Backbone alignment watch dominance · information · equilibrium
Start with the first third of Polak (dominance, backward induction, Nash). Pair with Coursera GT I for a tighter, formal pass.

2. Nash equilibrium — fixed points inside someone else’s game

Nash equilibrium: given everyone else’s strategy, no one benefits from unilateral deviation. Existence is guaranteed (mixed strategies) for finite games, but selection is the real war: multiple equilibria, focal points, and risk dominance.

Refinements (Selten) subgame perfection · trembling-hand perfection
  • Subgame perfect: removes non-credible threats in sequential games.
  • Trembling-hand perfect: robust to small mistakes.
Sovereign vs Synthetic read: equilibrium is “locally stable” — not “good.” Low-sovereignty equilibria can be risk-dominant. Your job is to change the game so high-sovereignty equilibria become both payoff- and risk-dominant for bounded agents.
Core Nash + dynamics course Nash · mixed · sequential
Primary: Polak. Secondary: Coursera GT I (formal reinforcement).
Bounded rationality (internal critique) assumption break Selten
When equilibrium concepts collide with real cognition.

3. Correlated equilibrium — narratives as coordination devices

Correlated equilibrium (CE): players condition on a shared signal (a “recommendation”). CEs can dominate Nash in welfare and can arise from adaptive learning.

In live systems, correlation devices are the “obvious” signals: alerts, guidelines, feeds, authority dashboards. Whoever owns the correlation device largely owns equilibrium selection.
Repeated interaction + correlation primary Aumann
Aumann’s repeated-games lens is the clean bridge into “norms as equilibrium.”
Information design (propaganda as mechanism) weaponized signaling Kamenica · Gentzkow
“Bayesian persuasion” formalizes how a sender chooses a signal structure to shape beliefs and actions.

4. Global games — thresholds, fear, and why people don’t move

Global-games logic: in coordination dilemmas with noisy private signals, multiplicity collapses into a unique equilibrium. Control the dispersion of belief and you prevent coordinated exits.

Key mechanism lever: convert suspicion into private dread instead of common knowledge. Exit fails because agents can’t infer whether others will move.
Where to learn the prerequisite machinery prereq Bayesian games
Bayesian games (types) are the prerequisite lens for global-games reasoning.

5. Repeated games — norms and the war on the future

Repeated games: outcomes can be sustained by credible future punishments if players are patient, can monitor, and can commit to punish.

Four enablers patience · monitoring · credible punishment · renegotiation-proofness
  1. Patience (low discount rates)
  2. Monitoring (observability)
  3. Credible sanctions (punishment actually executed)
  4. Renegotiation control (no immediate incentive to dissolve sanctions)
Synthetic pattern: raise discount rates (precarity/inflation), fragment observability (feeds/overload), neutralize punishment (amnesty for elites, crush whistleblowers), then demand “unity” to dissolve sanctions.
Formal frame primary Aumann
Repeated games and macro applications.
Emergence of cooperation primary Axelrod
Iterated Prisoner’s Dilemma tournaments; reciprocity conditions.

6. Signaling — costly proof vs engineered appearances

Signaling games: sender has private type, chooses a signal, receiver updates beliefs and acts. Equilibria: separating, pooling, semi-separating. Off-path beliefs decide what “unexpected integrity” means.

Design axis: make real signals costly in the right way (time/skill/patience), and make mimicry expensive for false types over long horizons.
Canonical model paper Spence
Separating vs pooling; credentialism as signaling.
Information structure as weapon information design Kamenica · Gentzkow
Choose signals to shape posteriors and actions: “persuasion as mechanism.”

7. Bargaining — disagreement payoffs and the price of escape

Bargaining is about the disagreement point (outside option). If your disagreement payoff is near zero, you accept anything.

Mechanism-level translation: lowering the disagreement payoff is a direct engineering route to “rational compliance.”
Alternating-offers (canonical) paper Rubinstein
Patience (discount factors) becomes bargaining power.
Commitment & credible threats conflict Schelling
Bargaining power often equals commitment power.

8. Auctions — when access and visibility are secretly sold

Auctions allocate scarce goods; platforms often turn attention, ranking, bandwidth, and priority into auctioned commodities — sometimes with opaque rules, asymmetries, and hidden quality scores.

Visibility: identify where auction logic is present even when no one says “auction.”
Revenue-optimal auctions lecture Myerson
Mechanism design meets auctions; the blueprint behind modern ad auctions.
Engineering view chapter Nisan
Mechanisms as code; incentive compatibility and VCG in CS language.

9. Mechanism design — the real battleground

Mechanism design asks: given a social choice rule, what message spaces + outcome rules make “truth-telling / rule-following” an equilibrium?

Impossibility & trade-offs Arrow · Gibbard–Satterthwaite · Myerson–Satterthwaite
  • Arrow: no perfect voting rule under standard fairness axioms (≥3 options).
  • Gibbard–Satterthwaite: strategy-proofness is basically impossible for general social choice (≥3 options) without dictatorship or restrictions.
  • Myerson–Satterthwaite: efficient, budget-balanced, individually rational bilateral trade fails in general with private information.

Implication: sovereign design chooses which failures are acceptable — explicitly.

Commitment problem: central designers can rewrite rules mid-stream. Sovereign mechanisms must minimize reliance on central commitment and maximize exit/fork capacity.
Foundations Nobel lectures Hurwicz · Maskin · Myerson
Mechanism design as institution engineering; implementation theory; optimal mechanisms.
Mechanisms as code (platform era) AGT Nisan · Roughgarden
Computational constraints + learning dynamics + large-scale environments.

10. Incomplete contracts — emergency clauses and residual control

Incomplete contracts: you can’t specify every contingency. What matters is who holds residual control rights when ambiguity arrives.

Synthetic pattern: “emergency” is the catch-all placeholder for residual power.
Residual control (core reference) Nobel lecture Hart
Ownership and control when contracts can’t fully specify the future.

11. Evolutionary games — strategy as meme, not plan

Evolutionary game theory replaces perfect calculation with strategies that reproduce proportional to fitness. ESS: once common, no mutant can invade.

Ecology is a mechanism: change the payoff environment and you change which strategies survive.
ESS foundation book Maynard Smith
The canonical ESS text; population games as strategic environments.
Fairness as equilibrium paper / lecture Binmore
Fairness norms as stable outcomes of repeated bargaining and evolutionary pressure.
Integrated evolutionary + experimental book Gintis
Bridges classical, evolutionary, and empirical game theory; useful for “humans aren’t utility bots.”

12. Network games & topology — graph structure as destiny

Agents exist on graphs. Topology determines observability, influence, and capture: hubs are leverage points. Sovereign topology targets modularity, redundancy, and reroutability.

Network games + platform mechanisms converge in algorithmic game theory: large-scale systems where the “game” is literally code.
Platform-era toolkit AGT Roughgarden Price of Anarchy
Congestion games, price of anarchy, auctions, learning dynamics — mechanisms in networks.

13. Behavioral, psychological, and intra-agent games

Real agents: loss aversion, present bias, status/shame dynamics, cognitive overload. Intra-agent games: present self vs future self (commitment devices).

Synthetic exploitation often targets: loss aversion (“you will lose X”), status (“good people do Y”), and time horizon (precarity + dopamine loops).
Bounded rationality map Nobel lecture Kahneman
Heuristics and biases; where expected-utility assumptions break.
Bounded rationality from inside GT internal critique Selten
“Good decisions without utility/probability judgments.”

14. Algorithmic game theory — computational asymmetry

AGT: mechanisms and equilibria implemented as algorithms on massive networks. Core realities: computing equilibria can be hard; learning dynamics matter; designers can adapt the game in real time.

Design constraint: keep strategies cognitively simple; make verification cheap; push cheating/surveillance into expensive computation.
Core course + notes video/notes Roughgarden
Price of anarchy, Myerson’s Lemma, auctions, no-regret learning.
Reference compendium book PDF Nisan · Roughgarden et al.
Algorithmic Game Theory (edited volume).

15. Sovereign-Stack design constraints (derived)

15.1 Accept impossibility & trade-offs no perfect mechanism exists

Choose where you tolerate inefficiency vs manipulability vs restrictions. Pretending you can have “fair + efficient + strategy-proof + voluntary + universal” is itself a control story.

15.2 No blind trust in central commitment assume rule patching

Prefer cryptographic enforcement, local enforceability, and low switching costs. Make rule change slow, visible, consent-based, and forkable.

15.3 Tame the designer problem mechanism design is power

Distribute design capacity; encode fork norms; include kill-switches and succession pathways.

15.4 Robust simplicity monk-simple rules

Small rule sets that degrade gracefully; multiple overlapping simple mechanisms instead of one “clever” fragile one.

15.5 Topology-aware design modular · redundant · reroutable

Avoid single points of failure (technical or social). Hubs may exist, but must be contestable and replaceable.

15.6 Cognitive & time constraints agents are overloaded

Design defaults (what happens if someone does nothing). Reduce required global knowledge. Treat “simplicity under stress” as a first-class constraint.

15.7 Explicitly model null outcomes fragmentation / predation

Design to reduce severity of failure modes, not only optimize best-case outcomes.

Meta-critique (how GT became governance) documentary Adam Curtis
Track the migration of game-theoretic “human models” into policy and management.

16. Synthesis

Game theory: how behavior stabilizes under rules + beliefs. Mechanism design: how to choose rules so certain patterns become self-enforcing.

Design target: for a bounded, partially manipulated agent in a hostile environment, the safest “rational” move must still align with property, voluntary contract, long-run integrity, and exit capacity.

The perfected move is not to win the old game. It is to build mechanisms where sovereignty and rationality converge — and remain alive under attack, error, and time.

Minimal “Hard Path” Through the Canon

Finite path that hits Nash, repeated games, signaling, bargaining, auctions, mechanism design, behavioral corrections, AGT/platform layer, and meta-critique.

Resource Index (linked)

All items below have direct links. Replace / expand as needed.

Ben Polak — Game Theory (ECON 159) course
Dominance, backward induction, Nash, repeated games, signaling, bargaining, auctions.
Game Theory I — Jackson / Leyton-Brown / Shoham course
Compact formal pass: Nash, mixed strategies, sequential play, basic tools.
Game Theory II — Jackson / Leyton-Brown / Shoham course
Extends to Bayesian games, auctions, and mechanism design foundations.
Tim Roughgarden — Algorithmic Game Theory (CS364A) course platforms
Price of anarchy, network games, auctions, learning/no-regret dynamics, Myerson’s Lemma.
von Neumann & Morgenstern — Theory of Games and Economic Behavior book
Foundational origin of formal game theory in economics.
John von Neumann: Mathematician (1966) film
Cold War–era cultural artifact: the atmosphere that surrounded early strategic modeling.
John Harsanyi — Games with Incomplete Information (Nobel Lecture) PDF
Types + Bayesian games: the prerequisite machinery for signaling, auctions, and global-games logic.
Robert Aumann — Repeated Games & Applications (Nobel Lecture) PDF
Formal repeated-game lens for norms, credibility, and long-run cooperation.
Michael Spence — Job Market Signaling (QJE, 1973) paper
Separating vs pooling equilibria; canonical costly-signal model.
Kamenica & Gentzkow — Bayesian Persuasion (AER, 2011) information design
Formal language for engineered narratives: choose signal structures to shape posteriors and actions.
Ariel Rubinstein — Bargaining (Alternating Offers) paper
Discount factors (patience) translate into bargaining power; canonical model.
Thomas Schelling — Commitment & Conflict (Nobel Lecture) strategy
Focal points, credible threats, brinkmanship: bargaining power as commitment power.
Noam Nisan — Introduction to Mechanism Design (for CS) chapter
Mechanism design as engineering: incentive compatibility, VCG, classic impossibilities.
Leonid Hurwicz — Mechanism Design (Nobel Lecture) PDF
Foundational framing: institutions as mechanisms; align “guardians” with incentives.
Eric Maskin — Implementation Theory (Nobel Lecture) PDF
How social choice rules can be implemented via mechanisms under strategic behavior.
Roger Myerson — Optimal Mechanisms & Auctions (Nobel Lecture) PDF
Mechanism architecture + auction design; core blueprint behind modern platform auctions.
Oliver Hart — Incomplete Contracts & Control (Nobel Lecture) lecture
Residual control rights decide reality when the contract runs out of specified contingencies.
Robert Axelrod — The Evolution of Cooperation book
Iterated PD tournaments; reciprocity conditions; emergence of cooperation.
John Maynard Smith — Evolution and the Theory of Games book ESS
Canonical ESS reference; evolutionary stability as a solution concept.
Ken Binmore — How and Why Did Fairness Norms Evolve? PDF fairness
Fairness as equilibrium selection under repeated interaction and bargaining constraints.
Lloyd Shapley — A Value for N-Person Games PDF cooperative GT
Axiomatic payoff division; Shapley value foundation.
Daniel Kahneman — Maps of Bounded Rationality (Nobel Lecture) behavioral PDF
Heuristics/biases and prospect-theory lens: where classical rationality breaks.
Reinhard Selten — Bounded Rationality (Lindau lecture) assumptions
“Good decisions without utility & probability judgments.”
Algorithmic Game Theory (edited volume) book PDF Nisan · Roughgarden et al.
Standard reference: algorithms for equilibria, auctions, mechanism design, price of anarchy.
The Trap: What Happened to Our Dream of Freedom (Adam Curtis) documentary meta-critique
How game-theoretic “human models” migrated into governance/management as hidden architecture.
Ariel Rubinstein on Rationality & Utility Theory (EconTalk) meta-critique podcast
Internal critique: models as language vs literal prediction engine.
Frontiers for Young Minds — “Game Theory—More Than Just Games” article
Lightweight explainer useful for fast refresh or onboarding a new reader.
Herbert Gintis — Game Theory Evolving PDF/excerpt evolutionary · experimental
Integrated approach: strategic interaction grounded in actual human behavior and evolutionary dynamics.

Tip: keep all external links in this page (module) and keep your “curriculum home” page as link-only navigation. That makes swapping resources later frictionless.