The Paradigm Shift: From Rational Agent to Bounded Rationality

The edifice of neoclassical economics rests upon a singular, elegant assumption: that individuals behave as rational agents, systematically maximizing utility subject to constraints. This *homo economicus*—a creature of consistent preferences, flawless calculation, and unwavering self-interest—provided the axiomatic bedrock for a century of theoretical development, from Walrasian general equilibrium to the efficient market hypothesis. The model’s beauty lies in its parsimony; by reducing decision-making to a problem of constrained optimization, it rendered human behavior mathematically tractable and theoretically pristine. Yet this very elegance sowed the seeds of its own undoing. For decades, the discipline operated under a tacit agreement: deviations from the rational model were dismissed as “noise,” irrelevant anomalies that would wash out in aggregate or reflect temporary irrationality. The laboratory, however, had other plans. A cascade of experimental findings, meticulously documented from the 1950s onward, revealed that these deviations were not random errors but systematic, predictable patterns—the “hidden currents” of human cognition. This chapter traces the intellectual archaeology of that paradigm shift, moving from the crystalline certainties of expected utility theory to the textured landscape of bounded rationality, where Herbert Simon’s satisficing agent gave way to Daniel Kahneman and Amos Tversky’s prospect theory. It argues that anchoring, loss aversion, and framing are not curiosities to be explained away but central features of a new descriptive science of decision-making, one that acknowledges the architecture of the mind as fundamentally ecological—adapted to real-world complexity through heuristics that are efficient, if sometimes erroneous.

The Homo Economicus: Foundations of the Rationalist Paradigm

The conception of the rational agent did not emerge ex nihilo; it was the culmination of a long intellectual project to mathematize economic behavior. The classical economists—Adam Smith, David Ricardo, John Stuart Mill—spoke of the “economic man” in more qualitative terms, emphasizing self-interest and calculation but without the formal apparatus of modern utility theory. The decisive leap occurred in the late 19th and early 20th centuries with the marginalist revolution, particularly through the work of William Stanley Jevons, Carl Menger, and Léon Walras, who introduced the notion of utility as a quantifiable, cardinal magnitude. This set the stage for the ordinalist revolution of the 1930s and 1940s, led by John Hicks and Roy Allen, which dispensed with cardinal utility in favor of a preference ordering approach. Yet the core postulate of rational maximization remained intact. The individual was assumed to possess a complete and transitive preference relation over all conceivable outcomes, to hold correct probabilistic beliefs (or subjective probabilities conforming to the axioms of probability), and to choose the option that maximized expected utility.

This framework was crystallized in the von Neumann-Morgenstern (1944) expected utility theory (EUT), which provided an axiomatic derivation of utility maximization from a set of seemingly plausible rationality postulates: completeness, transitivity, continuity, and independence. The independence axiom, in particular, became the theoretical linchpin. It states that if an agent prefers lottery $L_1$ to $L_2$, then for any third lottery $L_3$ and any probability $0 < p < 1$, the agent must prefer the mixed lottery $pL_1 + (1-p)L_3$ to $pL_2 + (1-p)L_3$. This axiom rules out certain types of risk attitudes, ensuring that preferences are linear in probabilities and that utility is a function of final wealth states alone. Under EUT, the decision-maker’s problem is to compute the expected utility $EU = \sum_{i} p_i u(x_i)$ for each option and select the one with the highest value. The model is prescriptive in its elegance: it tells the agent how they *should* decide if they are to be coherent.

The empirical implication of this paradigm is that observed choices should satisfy the axioms. Any systematic violation is not merely a mistake but a logical contradiction, a sign of irrationality. For decades, economists treated such violations as either the product of ignorance (which education could correct) or of externalities (like emotions or social pressures) that could be abstracted away in the idealized market. The rational agent was a *theoretical primitive*, not a psychological claim. As Milton Friedman (1953) argued in his famous essay on positive economics, the realism of assumptions mattered less than the predictive accuracy of the model; even if individuals did not consciously maximize, they would behave *as if* they did, thanks to evolutionary or market selection pressures. This “as-if” defense insulated the paradigm from direct psychological challenge, creating a protective moat between economics and the messy realities of human cognition.

Yet the moat was about to be breached—not by philosophical critique alone, but by controlled experiments that placed the axioms under the microscope of the laboratory. The first fissures appeared in the work of the pioneer of behavioral decision theory, Ward Edwards (1954), who demonstrated that people’s subjective probability weighting often violated the linear weighting implied by EUT. But the truly damning evidence came from two paradoxes that would become canonical in the literature: the Allais paradox (1953) and the Ellsberg paradox (1961). Maurice Allais, a French economist and Nobel laureate, presented subjects with a choice between two pairs of lotteries. In the first pair, most people preferred a certain $1 million over a 10% chance at $5 million and an 89% chance at $1 million (the “certainty effect”). In the second pair, the same people often preferred an 11% chance at $1 million and an 88% chance at nothing over a 10% chance at $5 million and a 90% chance at nothing. These choices violate the independence axiom because the common outcome (the 89% or 90% chance at $1 million) should cancel out, leaving the comparison between the 10% chance at $5 million and the 1% chance at $1 million—yet preferences reversed. Daniel Ellsberg’s urn paradox similarly violated the assumption of known probabilities: faced with an urn containing 90 red and 10 black balls, people preferred to bet on “red” rather than “red or black” (which has probability 1), and also preferred to bet on “yellow” (10 balls of another color) rather than “black or yellow.” This ambiguity aversion—preferring known risks over unknown ones—contradicted the Savage (1954) extension of EUT to subjective probabilities, which required that beliefs be represented by a single, coherent probability measure.

These paradoxes were not mere footnotes; they were logical impossibilities within the standard model. If the axioms were truly descriptive of rational choice, no coherent individual could exhibit such patterns. Yet they were robust across cultures, contexts, and even among economists themselves (who, when tested, often violated the axioms they taught). The rational agent model was not just psychologically unrealistic; it was *logically* unable to accommodate observed behavior without ad hoc modifications. The stage was now set for a fundamental rethinking of the decision-making process—one that would abandon the quest for a universal, context-independent rationality and instead seek to understand the heuristics and biases that actually govern human judgment.

Herbert Simon's Challenge: Bounded Rationality and Satisficing

While the paradoxes of Allais and Ellsberg provided empirical ammunition, the theoretical groundwork for an alternative had been laid a decade earlier by Herbert A. Simon, a polymath whose work spanned economics, psychology, computer science, and organizational theory. In his seminal 1955 paper “A Behavioral Model of Rational Choice” and his earlier