The Revolution Against Rationality: From Homo Economicus to Bounded Rationality

Imagine standing at a crossroads, not of geography but of decision-making. In your wallet is a crisp $100 bill. You are offered a gamble: a 50% chance to win an additional $200, and a 50% chance to win nothing. The expected value of the gamble is $100—exactly what you already have. A purely rational economic agent, the *Homo economicus* of classical theory, would be indifferent. Yet, most people faced with this choice feel a strong pull toward the sure $100. They are *risk-averse*. Now, consider the mirror image: you have lost $100 and are offered a gamble to get it back—a 50% chance to win $200 and a 50% chance to win nothing. Here, most people leap at the gamble, displaying *risk-seeking* behavior. The same expected value, the same objective parameters, but diametrically opposite choices depending on whether the scenario is framed as a gain or a loss. These are not anecdotes; they are replicable, systematic patterns that expose a profound disconnect between the idealized model of human rationality and the messy, contextual reality of human cognition. This chapter chronicles the revolution that overthrew the tyranny of *Homo economicus* and established the science of bounded rationality. It is the story of how economics, the quintessential social science of choice, was forced to confront the very creature it sought to model: the real, fallible, pattern-driven human mind.

The Unyielding Pillar: The Neoclassical Model of *Homo Economicus*

For over a century, the bedrock of mainstream economics rested on a singular, elegant, and profoundly influential conception of humanity: *Homo economicus*. This theoretical construct was not a description of any actual person but a necessary abstraction, a "as-if" model designed to render the chaotic social world tractable to mathematical analysis. At its core were a set of razor-sharp, interlocking assumptions about human cognition and motivation.

First, perfect information. The rational agent is presumed to know all relevant alternatives, their precise probabilities, and their full consequences. There is no cost to acquiring this information; it is simply given. Second, unbounded computational capacity. Given that information, the agent can effortlessly process it, performing complex probabilistic calculations and utility comparisons without fatigue, error, or cognitive strain. Third, consistent and stable preferences. These preferences are *complete* (the agent can rank any two options) and *transitive* (if A is preferred to B, and B to C, then A is preferred to C). This creates a coherent, well-ordered preference hierarchy. Fourth, utility maximization. The agent’s sole driving force is to choose the option that yields the highest possible *expected utility*—a cardinal measure of satisfaction derived from multiplying the utility of each outcome by its probability and summing across all possibilities. This process is not just a description of behavior; it is a normative prescription for how one *should* behave to be rational.

This model, formalized in the expected utility theory (EUT) of Daniel Bernoulli (1738) and later integrated into the general equilibrium frameworks of Walras, Marshall, and Pareto, provided an immensely powerful tool. It allowed economists to prove elegant theorems about market efficiency, the optimality of competitive equilibria, and the deterministic effects of policy changes. The *Homo economicus* was a ghost in the economic machine, a logical placeholder whose consistent maximizing behavior generated stable, predictable outcomes at the aggregate level. The model’s beauty lay in its parsimony and its ability to generate clear, testable predictions about prices, quantities, and resource allocation.

Yet, this very elegance was also its Achilles' heel. The assumptions were not mild approximations; they were heroic, often flagrantly contradicted by casual observation. Do people truly know all their options? Can they truly perform Bayesian updating on the fly? Are preferences truly stable and transitive? The model’s defenders argued that its purpose was not to describe psychology but to predict market outcomes, and that any systematic individual errors would cancel out in the aggregate ("the law of large numbers" defense). But a growing cadre of scholars, initially from psychology and later from within economics itself, began to suspect that these "errors" were not random noise. They were *systematic*, *predictable*, and *fundamental* to the architecture of human thought. The first cracks in the edifice appeared not from abstract critique, but from concrete, deceptively simple experiments.

The First Cracks: Allais and Ellsberg Reveal the Cracks in Expected Utility

The peace of the *Homo economicus* world was shattered by two experiments, conducted in the early 1950s, that exposed fatal inconsistencies in the expected utility framework. They did not merely show that people make mistakes; they demonstrated that people *systematically* violate the very axioms—like the *independence axiom* and the *sure-thing principle*—upon which EUT is built.

The Allais Paradox: The Violation of Independence

Maurice Allais, a French economist and later Nobel laureate, designed a thought experiment (later validated in laboratory settings) that targeted the independence axiom. This axiom states that if an agent prefers option A over option B, they must also prefer a probabilistic mixture of A with any third option C over the same mixture of B with C. In simpler terms, a common consequence or "common component" should not affect the relative preference between two options.

Consider the following two choice problems:

Problem 1:

• **Option A:** A sure gain of $1 million.
• **Option B:** 10% chance to gain $5 million, 89% chance to gain $1 million, 1% chance to gain nothing.

Problem 2:

• **Option C:** 11% chance to gain $1 million, 89% chance to gain nothing.
• **Option D:** 10% chance to gain $5 million, 90% chance to gain nothing.

Logically, Problem 2 is simply Problem 1 with the sure $1 million (the "common consequence") subtracted from both options. According to the independence axiom, a rational EUT maximizer who prefers A over B in Problem 1 must also prefer C over D in Problem 2. Why? Because the common $1 million (present in A and B with 89% and 100% probability respectively) is irrelevant to the comparison between the *additional* risks and rewards.

The experimental results were striking and robust. Approximately 82% of subjects chose the sure $1 million (A) in Problem 1, displaying risk aversion for large gains. But in Problem 2, 83% of those same individuals chose the riskier Option D (10% at $5M) over the smaller-chance sure gain (C). They reversed their preference solely because the sure option was no longer present as a "common consequence." This is a direct violation of the independence axiom. The presence of the safe, large-sum option in Problem 1 made the risky B seem less attractive by comparison (a phenomenon later understood as *certainty effect* or *pseudocertainty*), an influence that should have been irrelevant under EUT. Allais had proven that human choice is context-dependent in ways the model could not accommodate.

The Ellsberg Paradox: The Fear of the Unknown

While Allais attacked the mathematics of risk (known probabilities), Daniel Ellsberg (1961) targeted the very definition of probability itself, distinguishing between *risk* (known probabilities) and *ambiguity* (unknown probabilities). The expected utility model treats both identically; a rational agent should be indifferent between a bet on a known urn composition and a bet on an unknown one, provided the perceived probabilities are identical.

Ellsberg’s classic experiment involves two urns:

• **Urn A:** Contains 50 red balls and 50 black balls. Probabilities are known.
• **Urn B:** Contains 100 balls that are either red or black, but the exact number is unknown. The pro