Game theory is the mathematical study of strategic interaction — how rational agents make decisions when their outcomes depend on each other's choices. Founded by John von Neumann and Oskar Morgenstern in 1944 and extended by John Nash in the 1950s, it provides a formal language for competition, cooperation, negotiation, and mechanism design.
## Core concepts
**Nash equilibrium** is the central solution concept: a combination of strategies where no player can improve their outcome by unilaterally changing their choice, given what the other players are doing. In a Nash equilibrium, every player is playing their best response to everyone else. Real-world examples: competitive pricing between duopolists; bidding strategies in procurement auctions; network congestion (Braess's paradox).
**Zero-sum vs non-zero-sum**: in a zero-sum game, one player's gain is exactly another's loss (chess, poker). In non-zero-sum games, outcomes can be mutually beneficial (trade agreements) or mutually harmful (arms races). Most business situations are non-zero-sum — the key insight is finding strategies that capture more of the pie while the pie itself grows.
**Mechanism design** (inverse game theory): design the rules of a game so that self-interested agents, acting rationally, produce the socially desired outcome. Applications: eBay's auction design, FCC spectrum auctions, organ-donation matching markets, carbon trading systems.
## Game theory in AI
Modern AI systems encounter game-theoretic settings in several ways:
- **Multi-agent reinforcement learning**: when multiple AI agents interact in an environment (robotics, autonomous vehicles, trading), Nash equilibria define stable joint strategies. Training agents to find equilibria is computationally hard but tractable in specific settings.
- **Adversarial machine learning**: the game between a model and an adversary (trying to fool it with adversarial examples) is formally a minimax game. GANs (Generative Adversarial Networks) are a direct application of zero-sum game theory — the generator and discriminator play a minimax game until equilibrium.
- **AI alignment**: mechanism design informs how to structure AI system incentives so that self-interested optimization produces aligned behavior.
- **Auction and marketplace design**: large e-commerce platforms (Alibaba, Shopee, Lazada) and advertising systems run thousands of auctions per second, designed on game-theoretic principles.
## Practical relevance for enterprise teams
Enterprise AI teams rarely write game-theoretic proofs, but the intuitions matter:
- When deploying AI in competitive contexts (pricing bots, bid optimizers), expect competitors to adapt. The Nash equilibrium is what you eventually converge to — model it explicitly rather than discovering it expensively in production.
- When designing AI-driven internal platforms (resource allocation, internal carbon credits, shared compute budgets), mechanism design principles prevent gaming and ensure the system's rules create the incentives you actually want.
Where AIMenta applies this
Service lines where this concept becomes a deliverable for clients.
Beyond this term
Where this concept ships in practice.
Encyclopedia entries name the moving parts. The links below show where AIMenta turns these concepts into engagements — across service pillars, industry verticals, and Asian markets.
Other service pillars
By industry