Rationality across econ + psychology + philosophy

Start with the surprise: in economics, “rational” does not mean “smart,” “selfish,” or “cold.” It means consistent in a precise technical sense. An agent has preferences over outcomes; those preferences are complete (any two outcomes can be compared) and transitive (if you prefer $x$ to $y$ and $y$ to $z$, you prefer $x$ to $z$). Add the weak axiom of revealed preference — if you pick $x$ when $y$ was available, you don’t turn around and pick $y$ when $x$ is still on offer — and you have the procedural core. Nothing here says anything about what you should want. It only says your choices hang together.

From consistency to expected utility. The famous result — von Neumann and Morgenstern (1944), extended to subjective probability by Savage (1954) — is a representation theorem. If your preferences over risky prospects satisfy four axioms (completeness, transitivity, continuity, and independence), then there exists a utility function such that you behave as if you were maximizing its expected value. You do not need to be computing anything; the theorem says only that consistent choices look like expected-utility maximization from the outside.

Updating beliefs. The third pillar is Bayesian updating: a rational agent revises beliefs by combining a prior with the likelihood of the evidence, in accordance with Bayes’ rule. New information moves the posterior in a disciplined, reversible way. In macro, this same logic extends to expectations: Lucas and Sargent’s rational-expectations program assumes agents forecast using the true model of the economy and make no systematic errors — the apparatus reaching from individual choice all the way up to how a whole economy responds to policy.

What counts as a departure. The apparatus is precise enough to be broken in named ways. The Allais paradox (1953) shows people violating the independence axiom; the Ellsberg paradox (1961) shows them violating Savage’s sure-thing principle by preferring known odds to unknown ones; preference reversals (Grether and Plott, 1979) show people ranking $A$ over $B$ but paying more for $B$ — a direct WARP violation. These are not vague complaints that “people are irrational.” They are sharp, reproducible failures of specific axioms, which is exactly what makes them useful: the apparatus tells you precisely what was violated. What to do about the violation — treat it as a descriptive failure, an “as if” that still predicts well enough, or a reason to extend the model — is the question the next three stages turn on. The full intra-economics handling — whether the departures became parameter patches or replaced the apparatus — belongs to a forthcoming behavioral-economics walkthrough; here we only mark the departure and note that economics has a clear norm to depart from.

Want the formal home of the apparatus? Choice axioms, utility representation, and revealed preference live in Ch 11 §11.1–§11.2; the catalog of expected-utility violations — Allais, Ellsberg, preference reversals — is the demarcation surface in Ch 19 §19.1. For the intellectual lineage of the rational-actor tradition — the Friedman–Lucas counter-revolution and its game-theoretic roots — see History of Economic Thought Ch.10 (The counter-revolution).

Rationality as bounded by cognition. The starting point is Herbert Simon (1955): cognitive resources are scarce, the world is complex, and so agents satisfice — they search for an option that is good enough rather than computing the optimum. On this view the economist’s rational agent is an idealization that ignores the actual problem the mind is solving, which is to decide well under limits of time, memory, and attention. Rationality is not the absence of limits; it is what reasoning looks like once you take the limits seriously.

Heuristics and biases. Kahneman and Tversky turned Simon’s framing into a research program with teeth. Their prospect theory (1979) replaced expected utility with a value function defined over gains and losses relative to a reference point — not over final wealth — with losses looming larger than equivalent gains. People also lean on heuristics — representativeness, availability, anchoring — that work well on average but produce predictable errors. The catalog is not a list of human stupidities; it is a map of where actual choice diverges from the expected-utility prediction, and in which direction.

A word of caution the field imposed on itself: not every entry in the heuristics-and-biases canon survived. The Open Science Collaboration (2015) and Camerer and colleagues’ replication study (2016, Science) re-ran many landmark experiments; some effects shrank, some vanished, and the surviving canon is smaller and tighter than the 1990s version. That correction matters here because it tells you which deviations from economic rationality are robust enough to build a cross-discipline argument on.

Rationality as ecological fitness. Now the intra-psychology fork. Gerd Gigerenzer and the ABC research group reject the framing in which heuristics are workarounds for cognitive deficiency. On their account, fast-and-frugal heuristics — recognition, take-the-best — exploit the structure of the environment, and in many real settings they match or beat weighted-and-additive regression models. This is the “less-is-more” effect: ignoring information can improve inference when the ignored information is noisy. For Gigerenzer, a heuristic that achieves superior performance by fitting its environment is not a bias to be corrected. It is a different, and frequently better, kind of rationality — ecological rather than logical. This is not a footnote to Kahneman; it is a rival program with its own thirty-year empirical record.

And rationality is not intelligence. Keith Stanovich’s dual-process work (What Intelligence Tests Miss, 2009) adds a final twist: the fast, automatic System 1 and the slow, deliberative System 2 come apart from IQ. High-intelligence individuals are just as prone to System-1 biases on rationality-quotient tasks as anyone else. Being smart and being rational, in the psychologist’s sense, are measurably different things. The formal apparatus for prospect theory and bounded rationality lives in Ch 19 §19.2 and §19.5; the behavioral lineage inside economics’s own history of thought sits in History of Economic Thought Ch.13 (Behavioral economics) — though note that Gigerenzer, a psychologist, does not appear there at all, which is itself part of the cross-discipline picture.

Hume: reason serves, it does not set ends. Read carefully, Hume is not anti-rationalist. He is drawing a boundary. Reason is instrumental — it works out how to get what you want — and the thing that sets what you want is something else. Paired with his is/ought gap (no “ought” follows from purely descriptive “is” premises), this gives the foundational move of the whole stage: rationality has a means side that reason governs and an ends side that it does not, and the standard for each is a different kind of thing. The economist’s apparatus is squarely on the means side; it takes the ends (preferences) as given and never asks whether they are the right ones.

Davidson: charity makes interpretation possible. Donald Davidson’s principle of charity (Inquiries into Truth and Interpretation, 1984) is the apparatus move that re-reads the entire psychology stage. To interpret another agent’s beliefs and actions at all, we must assume they are mostly coherent and mostly responsive to their world; radical, thoroughgoing incoherence would make interpretation impossible, because we would have no foothold for assigning beliefs in the first place. Crucially, charity is an interpretive constraint, not an empirical finding — it is a condition on what can count as a description of an agent. The implication for Kahneman is sharp: if the heuristics-and-biases findings genuinely showed radical irrationality, they would not be interpretable as findings about choice at all. The fact that we can read them as systematic departures means the incoherence is bounded by definition.

Two rationalities: epistemic and practical. Contemporary philosophy splits the question the economic apparatus fuses. Epistemic rationality concerns belief — what should I believe given the evidence? — and is anchored by work like Timothy Williamson’s Knowledge and Its Limits (2000). Practical reason concerns action — what should I do given my goals and beliefs? — and is anchored by figures like Christine Korsgaard. The two have different normative structure: the canons of evidence are not the canons of agency. Expected-utility theory collapses both into a single maximization (a rational agent maximizes expected utility over actions given beliefs), and philosophy has never been convinced the collapse is licit.

The distinction that does the work: descriptive versus normative. In philosophy “rational” is a normative term — it picks out what an agent ought to believe or do. The economic apparatus is descriptive-procedural (here is what consistent choice looks like) yet is constantly used normatively (here is what people should do). The psychological apparatus is descriptive-empirical (here is what people actually do). The slide from “is” to “ought” — from a pattern of choice to a verdict about rationality — is where most of the cross-discipline friction lives, and the philosophical apparatus is the venue where that slide gets examined rather than performed by accident. Economics has its own brush with this question: the methodological debate over Friedman’s “as if” defense, where prediction is held to license assumptions that are admittedly not descriptions, is the one place the textbook touches normative-coherence ground (Economics Ch 1 §1.6, Positive vs. Normative Economics; the same F-twist commitment whose lineage runs through History of Economic Thought Ch.10).

The scope-of-application axis. Put the three rungs together and a pattern appears that no single stage could show. Economic-rationality is right for one job: modeling choice under risk in finite-information settings, where preferences are stable and outcomes well-defined. Psychological-rationality is right for a different job, in two flavors — the bounded line describes how actual decision-making departs from that model in field settings where the economic scope conditions fail, and the ecological line describes how heuristics achieve adaptive performance by exploiting environmental structure. Philosophical-rationality is right for a third job: the normative evaluation of whether belief and action cohere. Read each discipline inside its own scope and they mostly do not contradict each other at all. They are doing different work.

Where the apparent disagreement is verbal. Most of the popular cross-discipline fights are scope-confusions wearing the costume of substantive dispute. “Psychology proved economics wrong” treats field-setting descriptive claims as if they refuted in-scope procedural ones — but a description of how people deviate does not refute a framework for when choices are consistent; it describes a different scope. “Economists don’t understand human nature” treats a deliberate abstraction as a failed description — but the apparatus never claimed to describe cognition. “Philosophers play word games” treats the normative-conceptual layer as semantic noise — but that layer is what makes the descriptive/normative distinction legible at all. Strip the scope-confusion out and these disputes evaporate. That verbal dissolution is genuine, and it accounts for the bulk of the heat.

The three conflicts that survive. What is left after scope specification is small, sharp, and real. First: axiom-level preference reversals — Grether–Plott, Tversky, framing effects on identical lotteries — violate WARP in ways the rational-inattention and rational-addiction patches cannot absorb, and whether that counts as “irrational” requires a normative call philosophy is the right venue for but on which philosophy is itself split. Second: whether ecological-fitness is a kind of rationality — Gigerenzer says a heuristic that exploits environmental structure and outperforms the “rational” model has earned the label; the Kahneman tradition says adaptive efficiency is not the same as normative justification. Third: what normative coherence requires — Davidson-style charity says radical incoherence is unintelligible, constraining what can count as irrationality, while practical-reason theorists argue for narrower coherence that allows more substantive deviation. Each of these is a place where the disciplines genuinely disagree, the apparent verbal dispute is not the right read, and no single discipline can adjudicate.

What “fundamental agreement” would have required — and didn’t happen. For the three fields to truly converge, they would have needed a shared definition of “rational” accepted as the same concept, a shared normative criterion settling when descriptive deviation counts as irrationality rather than as a different rationality, and agreement on whether procedural-consistency, ecological-fitness, and normative-coherence are one axis or three. None of that occurred. What occurred instead is that each discipline locked in its own scope, and most of the public argument became a fight over which scope to apply — which is why it is mostly verbal, and why the residue that isn’t matters so much.