Kahneman and Tversky plan statistical-intuition research at Café Rimon

Kahneman and Tversky plan statistical-intuition research at Café Rimon was the formative research decision that converted an animated seminar exchange into one of the most consequential collaborations in modern psychology. At a lunch at Café Rimon in Jerusalem, Daniel Kahneman and Amos Tversky decided to investigate whether sophisticated researchers possessed a dependable intuitive grasp of statistical principles. Their answer would expose a gap between expertise and judgment: people who taught and used statistics could still be misled by small samples, weak evidence, and the coherence of an appealing result.

From Seminar Debate to Research Question

The plan grew out of a seminar at the Hebrew University of Jerusalem in 1969, where Amos Tversky discussed an ongoing program at the University of Michigan asking whether people were good intuitive statisticians. The comparison with intuitive grammar made the issue vivid. A four-year-old could follow grammatical rules without knowing that the rules existed, but it was unclear whether adults possessed a comparable intuitive feel for basic statistical principles. The seminar initially produced a qualified yes, then moved toward a qualified no.

The decisive insight was self-implicating. Kahneman and Tversky recognized that their own intuitions were deficient despite years of teaching and using statistics. They were too willing to accept findings based on inadequate evidence and were inclined to gather too few observations in their own research. The proposed study therefore did not begin by treating statistical error as a defect confined to novices; it asked whether expert knowledge itself failed to produce the expected intuitive discipline.

The Café Rimon Plan

That Friday, the two researchers met for lunch at Café Rimon, a favorite gathering place for bohemians and professors in Jerusalem, and planned a study of the statistical intuitions of sophisticated researchers. The setting mattered less as a picturesque backdrop than as the scene in which a conversational discovery acquired an empirical form. Their objective was to test whether their own susceptibility to small-sample error was shared by colleagues whose professional identities rested on statistical competence.

They prepared a survey built around realistic research scenarios. The questions concerned the likelihood that an original experimental result would be successfully replicated when the sample was small and the number of observations a fictitious graduate student should collect. Tversky administered the survey to expert participants at a meeting of the Society of Mathematical Psychology, among them the authors of two statistical textbooks. The design thus placed the problem directly before people expected to understand it best.

What the Study Revealed

The results were not merely a demonstration that ordinary people sometimes reason badly. The expert respondents, like Kahneman and Tversky themselves, greatly exaggerated the chance that an experimental result would replicate even when it rested on a small sample. They also gave poor advice about the number of observations required for reliable research. The conclusion was sharp: even statisticians were not good intuitive statisticians.

This finding belongs to a larger pattern in the book. Experts frequently construct judgments from information that comes readily to mind, while failing to represent what is absent, uncertain, or statistically decisive. Comparable evidence appears in later research on prediction. Philip Tetlock’s twenty-year study collected more than 80,000 forecasts from 284 people who earned their living commenting on political and economic trends; the experts performed worse than an equal-probability baseline, and even specialists were not significantly better than nonspecialists in regions they knew best. The Café Rimon project supplied an early version of the same challenge to expertise: training and professional status did not guarantee calibrated judgment.

The calibration problem is also visible in formal probability assessments. When judges construct subjective distributions by selecting percentile estimates, proper calibration requires the true values to fall below the stated X01 and X99 limits at the expected rates, leaving approximately 98 percent within the interval. Studies instead found that actual values fell outside such extreme bounds in about 30 percent of cases, revealing confidence intervals that were systematically too narrow. Anchoring helps explain the distortion: people begin with a best estimate and adjust insufficiently upward or downward, producing a distribution tighter than their knowledge warrants.

Consequences for Decision Theory

The statistical-intuition project helped establish the empirical attitude that would govern Kahneman and Tversky’s later work. Rather than asking only what a rational decision maker ought to do, they examined the choices people actually found immediately compelling and then tested those intuitions against statistical standards. Their later research on prospect theory followed this route. After five years of studying gambles, they published “Prospect Theory: An Analysis of Decision under Risk,” a descriptive theory designed to document systematic violations of the axioms of rational choice. It explained risky decisions through features such as reference points, gains, and losses rather than through wealth alone.

The same movement from abstract rationality to observed judgment shaped their critique of conventional economics. Richard Thaler’s distinction between Econs and Humans captured the contrast: Econs possess a comprehensive and consistent view of outcomes, while Humans are limited by the information available at a given moment and are vulnerable to WYSIATI, narrow framing, priming, and preference reversals. Mental accounts show how this limitation operates in everyday choices. A person who paid for a basketball ticket is more likely to drive forty miles through a blizzard than someone who received a ticket free, because the first person experiences staying home as closing a mental account with a more negative balance, even though the purchase price is already sunk.

The sunk-cost fallacy extends the same logic into organizations. A company that has already spent $50 million on a failing project may invest another $60 million rather than redirecting those resources to a more promising project. From the firm’s perspective, earlier expenditure should not determine the new decision, but closing the account can feel like admitting failure. The escalation may serve the executive’s personal interests by postponing the day of reckoning, creating an agency problem between the manager and the organization’s shareholders.

Lasting Significance

The Café Rimon decision was small in physical scale but large in intellectual consequence. Two researchers examined their own confident intuitions, discovered that statistical expertise had not protected them from error, and designed a survey to test whether the weakness was general. The study’s importance lay not in humiliating experts but in changing what counted as evidence about judgment: competence had to be measured by performance, replication, and calibration rather than inferred from fluency, training, or reputation.

That principle remained central to their broader program. Humans are not irrational in the strong sense of being impulsive or impervious to reasons, but they cannot consistently satisfy the coherence demanded by the rational-agent model. They often need assistance from sound procedures, institutions, and policies. The lunch at Café Rimon therefore represents more than the origin of a single survey. It marks the moment when a conversation about intuitive statistics became a method for studying the distance between what seems convincing and what the evidence can support.