How Incomplete Information Affects Game Theory Models

In game theory, assuming that all players have perfect knowledge of their opponents’ payoffs, preferences, and motives is rarely realistic. This article explores how introducing incomplete information fundamentally alters the design of game theory models, transforming simple strategic interactions into complex games of beliefs, signaling, and probability. We will examine the shift from standard Nash equilibria to Bayesian frameworks, the role of Harsanyi’s transformation, and how designers model dynamic interactions when players must update their beliefs mid-game.

The Shift from Complete to Incomplete Information

In a game of complete information, every player knows the rules, the possible strategies, and the exact payoffs of all other players. When designers introduce incomplete information, this certainty vanishes. Players no longer know the exact payoffs or motivations of their competitors.

To design a model under these conditions, theorists must redefine how players make decisions. Instead of choosing a strategy based on fixed outcomes, players must make choices based on expected utility, factoring in the probability of facing different types of opponents.

Harsanyi’s Transformation: Modeling “Types”

Before 1967, modeling incomplete information was mathematically chaotic. Economist John Harsanyi resolved this by introducing a framework that converts a game of incomplete information into a game of imperfect information.

In this design, a fictional player called “Nature” moves first. Nature randomly assigns a “type” to each player (for example, a “high-cost” or “low-cost” competitor). * Players know their own type but do not know the types of their opponents. * Players do, however, share a common prior belief—a probability distribution over which types Nature is likely to select.

By using Harsanyi’s transformation, game designers can mathematically structure uncertainty, allowing players to calculate risks using probability.

The Solution Concept: Bayesian Nash Equilibrium

In standard games, the Nash Equilibrium identifies steady states where no player has an incentive to deviate. In games with incomplete information, designers must use the Bayesian Nash Equilibrium (BNE).

Under a BNE, a strategy must specify an action for every possible type a player might be, even if they only end up being one type. A BNE is reached when every player chooses a strategy that maximizes their expected payoff, given their beliefs about other players’ types and the strategies those types would play.

Dynamic Games and Information Revelation

When games of incomplete information are played sequentially (over multiple rounds), the design becomes even more intricate. Actions taken in early rounds reveal information about a player’s hidden type. Designers use two primary concepts to model this flow of information:

1. Signaling

The informed player acts first. Their action can signal their true type to the uninformed player. A classic example is a company offering a warranty; only a high-quality manufacturer (type) can afford to offer a long warranty, signaling their type to the consumer.

2. Screening

The uninformed player acts first, setting up choices to force the informed player to reveal their type. An insurance company, for example, offers a choice between high-deductible/low-premium plans and low-deductible/high-premium plans to screen high-risk drivers from low-risk drivers.

In these dynamic models, the solution concept shifts to the Perfect Bayesian Equilibrium (PBE). This requires players to continuously update their beliefs using Bayes’ Rule as they observe the actions of others throughout the game.

Core Design Challenges of Incomplete Information

Designing models with incomplete information introduces several key challenges: