Skip to content

Rational Inattention

Economic choice with optimally selected partial information under an attention cost or capacity limit.

Version
v1 · 2026-09-28 · History
Domain-specific #
11679
Domain group
Social Sciences
Origin domain
Economics & Finance
Subdomains
Information Economics, Macroeconomics → Economics & Finance
Aliases
RI model, Rationally inattentive choice

Core Idea

Rational inattention treats information as a scarce input to economic choice. Instead of assuming the agent sees every relevant state perfectly, a model lets the agent choose a signal or attention allocation under a processing-capacity constraint or information cost and then choose an action. The resulting action may appear inferior under a full-information benchmark yet be optimal once acquiring better information is costly.

Sims's foundational formalization uses a finite Shannon information channel to restrict how observations can influence actions. A later empirical application by Joo estimates an information-cost discrete-choice model in a laundry-detergent market and evaluates the introduction of Tide Pods. Its finding that evaluation friction can offset benefits is a model-dependent result, not a universal verdict on new products or a direct measurement of individual attention budgets.

Structural Signature

Sig role-phrases:

  • Decision agent — An actor chooses in an uncertain environment with a payoff criterion. It is constitutive. Counterfactual: A non-choosing passive sensor is not the economic decision maker.
  • Uncertain state and options — Relevant circumstances and available actions are not fully known ex ante. It is constitutive. Counterfactual: A deterministic choice with no information question lacks the central friction.
  • Selective information channel — The agent chooses or operates through a signal about the state. It is constitutive. Counterfactual: Fixed arbitrary ignorance without information selection is not the rational-inattention mechanism.
  • Capacity or information cost — Processing precision is limited or penalized, often measured by Shannon information. It is constitutive. Counterfactual: Free complete observation would remove the proposed attention constraint.
  • Optimized action rule — The agent chooses information and response to improve expected payoff net of the constraint. It is constitutive. Counterfactual: A lapse that contradicts the agent's own stated optimization cannot be rationalized by label alone.
  • Observed or modeled choice consequence — Selective attention may yield delayed, noisy or partial-information actions relative to a full-information benchmark. It is central. Counterfactual: An estimated welfare change is conditional on model and identification assumptions.

What It Is Not

  • Not arbitrary distraction. The information limitation must enter a choice problem.
  • Not all bounded rationality. This model makes attention acquisition or channel capacity explicit.
  • Not direct mind-reading. Empirical applications infer latent processing through model assumptions.
  • Not a universal welfare prediction. Effects depend on payoffs, signal costs and option sets.
  • Closest near-miss. A shopper misses a price tag for no modeled reason and buys a costly product: without a choice of signal or attention constraint, this is simply an error, not evidence for the model.

Scope of Application

  • Macroeconomics. Model slow or noisy adjustment to economic signals.
  • Consumer choice. Estimate consideration and evaluation costs.
  • Pricing. Study what shocks firms attend to when setting prices.
  • Policy analysis. Compare model-dependent outcomes under changed information environments.

Clarity

Rational inattention means choosing how much and what information to process before acting, because attention has a cost or capacity bound. Sims models a finite Shannon channel; Joo estimates an attention-cost choice model for detergent purchases. A mistaken purchase alone does not demonstrate rational inattention, and the Tide Pods welfare finding is conditional on the fitted model.

Manages Complexity

The agent must solve a joint signal-and-action problem. A richer signal can improve the decision but consumes capacity; an enlarged option set can improve matches yet raise evaluation burden. Empirical work must separate attention frictions from changing tastes, search exposure and ordinary noise, so the model's latent process is not directly observed.

Abstract Reasoning

  1. State the uncertain environment, actions and payoffs.
  2. Define an information channel or acquisition cost.
  3. Choose a signal and response jointly under the constraint.
  4. Compare the resulting action with a full-information benchmark.
  5. Identify what data could separate attention from other frictions.
  6. Report model-dependent rather than universal welfare conclusions.

Knowledge Transfer

The scarce-information logic can inform psychology or machine learning analogies, but this node is the economic model of optimizing an information structure and action under explicit costs. A person casually overlooking a detail is only analogous until that mechanism is specified.

Examples

Canonical

Sims's 2003 construction represents an economic agent as an information-processing channel of finite Shannon capacity: action can depend on an uncertain observed state only through a bounded flow of information. A candidate signal and ensuing action are jointly evaluated against expected payoff. The example is a formal model, not an observation that every person or firm literally computes channel capacity.

Mapped back: Decision agent → Sims's optimizing modeled agent; Uncertain state and options → external random economic signals and possible actions; Selective information channel → chosen limited-rate signal from observations; Capacity or information cost → finite Shannon channel capacity; Optimized action rule → payoff-sensitive action conditioned on processed signal; Observed or modeled choice consequence → model-predicted partial-information response.

Applied / In Practice

Joo's 2023 empirical study fits a rational-inattention discrete-choice model to a laundry-detergent market with retail promotion variation. The analysis of Tide Pods' entry estimates that additional options can raise evaluation costs enough to outweigh match-value gains in the fitted model. It is an attested data application of the framework, not direct proof of a psychological channel inside every shopper or a universal harm from new products.

Mapped back: Decision agent → modeled detergent shoppers; Uncertain state and options → product attributes and alternatives including Tide Pods; Selective information channel → latent consumer consideration/information structure inferred by the model; Capacity or information cost → estimated evaluation friction for a larger choice set; Optimized action rule → fitted discrete-choice rule under information costs; Observed or modeled choice consequence → conditional welfare estimate from the paper.

Structural Tensions

T1 — Decision Precision versus Information Cost. More accurate signals can improve choice but consume scarce capacity or require effort.

Diagnostic: How much information is worth acquiring?

T2 — Simple Choice Set versus Option Diversity. More options may improve match value while increasing evaluation burden.

Diagnostic: When does one effect outweigh the other?

T3 — Behavioral Fit versus Identification. Flexible latent attention can explain choices yet requires credible restrictions to distinguish it from preferences or noise.

Diagnostic: What evidence separates these explanations?

Structural–Framed Character

Rational Inattention is mixed-framed. The mathematics of constrained information processing is structural, while its use as a theory of economic choice depends on what counts as an agent's objective, uncertainty and information cost. Evaluative weight: the model's optimum is relative to a specified payoff; a welfare conclusion requires further normative and empirical assumptions, not a verdict built into the phrase “rational.” Human-practice dependence: limited attention can be a real cognitive constraint, but the Sims channel, prior and utility are an analyst's representation of it; the model can be applied to nonhuman decision makers only with justified role mapping. Institutional origin: economics and information theory supplied this named framework, rather than discovering one observer-free object called a Shannon capacity inside every shopper. Vocabulary travel: states, signals, costs and policies travel mathematically; detergent demand and consumer welfare do not. Import versus recognition: another decision problem with an endogenous costly signal and optimized action can instantiate the model; calling any distraction “rational inattention” imports a metaphor without its constitutive optimization.

The portable skeleton is Optimization: a joint information-and-action policy is chosen for an objective under an information constraint or penalty. That prime supplies the cross-domain choice/constraint relation. Rational Inattention specializes it with a Shannon-style information structure and economic interpretation; Bounded Rationality's heuristic/stopping signature is not automatically present. Its character: a formal optimization pattern whose claimed behavioral and welfare meaning remains model- and domain-framed.

Structural Core vs. Domain Accent

This section decides why Rational Inattention is domain-specific rather than a prime.

What is skeletal (could lift toward a cross-domain prime). An agent chooses what evidence to acquire and what act to take, balancing the expected value of improved discrimination against the cost of obtaining information. Decision variables, objective and constraint are a form of Optimization. The same thin relation can be recognized in experimental design, sensing and computation. It does not, by itself, say whether a person literally has a finite Shannon channel or whether a fitted information cost measures cognition.

What is domain-bound. The Sims framework makes uncertainty over states, a signal structure, attention cost/capacity and payoff-maximizing action jointly operational in economic choice. The retail application estimates this structure from observed laundry-detergent choices and evaluates a counterfactual product introduction. Its Tide Pods welfare claim is conditional on the fitted model and comparison, not a direct reading of shoppers' mental channels. Strip out endogenous attention allocation or the economic payoff interpretation and the remaining decision task may still be constrained optimization but not this account of rational inattention. Conversely, ordinary missed details or fatigue do not qualify by name alone.

Why this does not clear the prime bar. The mathematical apparatus is portable, yet recognition of this named model requires the information-choice structure and a defensible mapping of states, signals, acts and costs. Outside economics, similar limits might be measured or modeled differently; importing the name without that structure is analogy. Nor is the entry simply the current Bounded Rationality prime: that node's heuristic search and stopping commitments need not occur in an optimally selected information policy. Optimization carries the widely traveling skeleton. The domain-specific entry carries a historically and empirically framed theory of information-limited choice.

  • Related prime: Optimization, not an asserted parent. A rational-inattention model optimizes an information-and-action policy under a cost or capacity constraint, but the named abstraction is an economic decision theory that also interprets uncertainty and attention; it is not simply an optimization problem as the prime's object-type signature requires.

  • Conceptual relation: Bounded Rationality. Its current heuristic-search and aspiration/stopping roles need not occur in a Sims-style optimizing channel model, so no strict edge is asserted.

Neighborhood in Abstraction Space

Rational Inattention sits in a crowded region of the domain-specific corpus (32nd percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Decisions Under Constraint & Commitment (9 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With

  • Random decision error. Tell: Lacks payoff-guided information choice.
  • Full-information rational choice. Tell: Assumes the relevant state is observed without this friction.
  • Fixed missing data. Tell: Information absence is exogenous rather than optimally selected.
  • Search cost alone. Tell: Can overlap, but need not model a finite information-processing channel or endogenous signal structure.

References