Robyn Dawes

Robyn Dawes appears in the book as a major figure in the search for more reliable judgment. His importance lies less in offering a portrait of personality than in changing the standard by which prediction is evaluated: a method need not be sophisticated, psychologically persuasive, or tailored to every case to be useful. A rough formula, provided that its ingredients have some validity and are combined consistently, may outperform the confident expert who sees a compelling story in each individual case.

The Case Against Expert Intuition

Dawes’s work enters a broader argument about the weakness of human prediction. Philip Tetlock’s twenty-year study collected more than 80,000 forecasts from 284 people who made their living commenting on political and economic trends. The experts performed worse than a simple equal-probability baseline, and even specialists were not significantly better than nonspecialists in the regions they knew best. Knowledge produced only a slight improvement, while the most knowledgeable forecasters often became less reliable because knowledge strengthened their illusion of skill and their overconfidence.

The same problem appears in financial forecasting. In a Duke University survey, corporate chief financial officers produced 11,600 predictions of the following year’s Standard & Poor’s index. The correlation between their estimates and the actual returns was slightly less than zero. Their confidence intervals were also far too narrow: 67 percent of outcomes fell outside intervals designed to contain only about 20 percent of surprises. Dawes’s contribution is significant against this background because it supplies an alternative to intuition rather than merely documenting intuition’s failures.

Equal Weights and the Robust Formula

In his famous article “The Robust Beauty of Improper Linear Models in Decision Making,” Dawes argued that the complex machinery of multiple regression often adds little or no value. Multiple regression finds an optimally weighted combination of predictors in an original sample, but that apparent precision can reflect accidents of sampling. When the formula is applied to new cases, those accidental weights may reduce its performance.

Dawes showed that a set of predictors with some validity can be placed on comparable scales and combined with equal weights. Such an “improper” model can predict new cases about as accurately as the statistically optimal formula, and equal weighting may even be superior because it is less vulnerable to the peculiarities of the original sample. The practical lesson is deliberately unsettling: useful algorithms can sometimes be built without extensive prior statistical research. Existing statistics, or even carefully chosen common-sense indicators, may be enough if they are combined mechanically rather than reshaped by the impressions of the moment.

A Memorable Rule for Marital Stability

Dawes illustrated the power of simple aggregation with a compact predictor of marital stability: the frequency of lovemaking minus the frequency of quarrels. The example is memorable precisely because it refuses the usual theatrical image of expertise. It does not claim to capture every marriage, nor does it offer a deep causal theory of intimacy. It combines two relevant observations in a transparent rule and leaves the result to be interpreted without pretending that a more elaborate calculation must be better.

The formula also embodies Dawes’s larger methodological preference for consistency. Human judges are easily drawn toward vivid cases and coherent explanations, but the information that feels most meaningful is not necessarily the information that predicts best. A simple rule protects the decision from the shifting emphasis, selective attention, and overconfidence that enter when an evaluator continually revises the weights according to personal impressions.

Independence, Aggregation, and Error

The book connects Dawes’s equal-weighting argument to a general principle: error should be decorrelated. When many observers estimate the number of pennies in a jar, individuals may greatly overestimate or underestimate the true number, yet the average judgment can be remarkably accurate because independent errors tend to cancel. The benefit disappears when observers share a bias or influence one another, since their mistakes become correlated and the effective sample shrinks.

This logic is also part of sound police procedure. Multiple witnesses are prevented from discussing an event before giving testimony, not only to prevent collusion but also to stop initially independent witnesses from copying one another’s errors. Dawes’s defense of formulas belongs to the same family of ideas. Independent pieces of evidence retain more information, while a human decision maker who lets one impression reshape the interpretation of every subsequent clue creates redundancy and amplifies a common mistake.

Significance for Decision Making

Dawes’s argument extends beyond academic prediction. The book presents equal-weight formulas as plausible tools for decisions about medical treatment, stock selection, and other consequential outcomes. In low-validity environments, where the relationship between available information and future results is weak, the disciplined regularity of an algorithm can be an advantage. A doctor, interviewer, or portfolio manager may feel that personal contact or specialized knowledge adds insight, yet the added impression can lower accuracy when it receives excessive weight.

His work therefore sharpens the book’s distinction between psychological confidence and predictive validity. A formula can be unimpressive while remaining dependable; an expert can be eloquent while adding noise. Dawes does not make simplicity magical or claim that every equal-weight rule is sound. The supported claim is narrower and more powerful: once predictors have some independent validity, simple and stable combination often competes successfully with elaborate models and can decisively outperform unaided expert judgment.