EVERYBODY'S AN EXPERT
Putting predictions to the test.
by LOUIS MENAND
Issue of 2005-10-05
Posted 2005-11-28
Prediction is one of the pleasures of life. Conversation would wither
without it. "It won't last. She'll dump him in a month." If you're wrong, no
one will call you on it, because being right or wrong isn't really the
point. The point is that you think he's not worthy of her, and the
prediction is just a way of enhancing your judgment with a pleasant
prevision of doom. Unless you're putting money on it, nothing is at stake
except your reputation for wisdom in matters of the heart. If a month goes
by and they're still together, the deadline can be extended without penalty.
"She'll leave him, trust me. It's only a matter of time." They get married:
"Funny things happen. You never know." You still weren't wrong. Either the
marriage is a bad one-you erred in the right direction-or you got beaten by
a low-probability outcome.
It is the somewhat gratifying lesson of Philip Tetlock's new book, "Expert
Political Judgment: How Good Is It? How Can We Know?" (Princeton; $35), that
people who make prediction their business-people who appear as experts on
television, get quoted in newspaper articles, advise governments and
businesses, and participate in punditry roundtables-are no better than the
rest of us. When they're wrong, they're rarely held accountable, and they
rarely admit it, either. They insist that they were just off on timing, or
blindsided by an improbable event, or almost right, or wrong for the right
reasons. They have the same repertoire of self-justifications that everyone
has, and are no more inclined than anyone else to revise their beliefs about
the way the world works, or ought to work, just because they made a mistake.
No one is paying you for your gratuitous opinions about other people, but
the experts are being paid, and Tetlock claims that the better known and
more frequently quoted they are, the less reliable their guesses about the
future are likely to be. The accuracy of an expert's predictions actually
has an inverse relationship to his or her self-confidence, renown, and,
beyond a certain point, depth of knowledge. People who follow current events
by reading the papers and newsmagazines regularly can guess what is likely
to happen about as accurately as the specialists whom the papers quote. Our
system of expertise is completely inside out: it rewards bad judgments over
good ones.
"Expert Political Judgment" is not a work of media criticism. Tetlock is a
psychologist-he teaches at Berkeley-and his conclusions are based on a
long-term study that he began twenty years ago. He picked two hundred and
eighty-four people who made their living "commenting or offering advice on
political and economic trends," and he started asking them to assess the
probability that various things would or would not come to pass, both in the
areas of the world in which they specialized and in areas about which they
were not expert. Would there be a nonviolent end to apartheid in South
Africa? Would Gorbachev be ousted in a coup? Would the United States go to
war in the Persian Gulf? Would Canada disintegrate? (Many experts believed
that it would, on the ground that Quebec would succeed in seceding.) And so
on. By the end of the study, in 2003, the experts had made 82,361 forecasts.
Tetlock also asked questions designed to determine how they reached their
judgments, how they reacted when their predictions proved to be wrong, how
they evaluated new information that did not support their views, and how
they assessed the probability that rival theories and predictions were
accurate.
Tetlock got a statistical handle on his task by putting most of the
forecasting questions into a "three possible futures" form. The respondents
were asked to rate the probability of three alternative outcomes: the
persistence of the status quo, more of something (political freedom,
economic growth), or less of something (repression, recession). And he
measured his experts on two dimensions: how good they were at guessing
probabilities (did all the things they said had an x per cent chance of
happening happen x per cent of the time?), and how accurate they were at
predicting specific outcomes. The results were unimpressive. On the first
scale, the experts performed worse than they would have if they had simply
assigned an equal probability to all three outcomes-if they had given each
possible future a thirty-three-per-cent chance of occurring. Human beings
who spend their lives studying the state of the world, in other words, are
poorer forecasters than dart-throwing monkeys, who would have distributed
their picks evenly over the three choices.
Tetlock also found that specialists are not significantly more reliable than
non-specialists in guessing what is going to happen in the region they
study. Knowing a little might make someone a more reliable forecaster, but
Tetlock found that knowing a lot can actually make a person less reliable.
"We reach the point of diminishing marginal predictive returns for knowledge
disconcertingly quickly," he reports. "In this age of academic
hyperspecialization, there is no reason for supposing that contributors to
top journals-distinguished political scientists, area study specialists,
economists, and so on-are any better than journalists or attentive readers
of the New York Times in 'reading' emerging situations." And the more famous
the forecaster the more overblown the forecasts. "Experts in demand,"
Tetlock says, "were more overconfident than their colleagues who eked out
existences far from the limelight."
People who are not experts in the psychology of expertise are likely (I
predict) to find Tetlock's results a surprise and a matter for concern. For
psychologists, though, nothing could be less surprising. "Expert Political
Judgment" is just one of more than a hundred studies that have pitted
experts against statistical or actuarial formulas, and in almost all of
those studies the people either do no better than the formulas or do worse.
In one study, college counsellors were given information about a group of
high-school students and asked to predict their freshman grades in college.
The counsellors had access to test scores, grades, the results of
personality and vocational tests, and personal statements from the students,
whom they were also permitted to interview. Predictions that were produced
by a formula using just test scores and grades were more accurate. There are
also many studies showing that expertise and experience do not make someone
a better reader of the evidence. In one, data from a test used to diagnose
brain damage were given to a group of clinical psychologists and their
secretaries. The psychologists' diagnoses were no better than the
secretaries'.
The experts' trouble in Tetlock's study is exactly the trouble that all
human beings have: we fall in love with our hunches, and we really, really
hate to be wrong. Tetlock describes an experiment that he witnessed thirty
years ago in a Yale classroom. A rat was put in a T-shaped maze. Food was
placed in either the right or the left transept of the T in a random
sequence such that, over the long run, the food was on the left sixty per
cent of the time and on the right forty per cent. Neither the students nor
(needless to say) the rat was told these frequencies. The students were
asked to predict on which side of the T the food would appear each time. The
rat eventually figured out that the food was on the left side more often
than the right, and it therefore nearly always went to the left, scoring
roughly sixty per cent-D, but a passing grade. The students looked for
patterns of left-right placement, and ended up scoring only fifty-two per
cent, an F. The rat, having no reputation to begin with, was not embarrassed
about being wrong two out of every five tries. But Yale students, who do
have reputations, searched for a hidden order in the sequence. They couldn't
deal with forty-per-cent error, so they ended up with almost fifty-per-cent
error.
The expert-prediction game is not much different. When television pundits
make predictions, the more ingenious their forecasts the greater their
cachet. An arresting new prediction means that the expert has discovered a
set of interlocking causes that no one else has spotted, and that could lead
to an outcome that the conventional wisdom is ignoring. On shows like "The
McLaughlin Group," these experts never lose their reputations, or their
jobs, because long shots are their business. More serious commentators
differ from the pundits only in the degree of showmanship. These serious
experts-the think tankers and area-studies professors-are not entirely out
to entertain, but they are a little out to entertain, and both their status
as experts and their appeal as performers require them to predict futures
that are not obvious to the viewer. The producer of the show does not want
you and me to sit there listening to an expert and thinking, I could have
said that. The expert also suffers from knowing too much: the more facts an
expert has, the more information is available to be enlisted in support of
his or her pet theories, and the more chains of causation he or she can find
beguiling. This helps explain why specialists fail to outguess
non-specialists. The odds tend to be with the obvious.
Tetlock's experts were also no different from the rest of us when it came to
learning from their mistakes. Most people tend to dismiss new information
that doesn't fit with what they already believe. Tetlock found that his
experts used a double standard: they were much tougher in assessing the
validity of information that undercut their theory than they were in
crediting information that supported it. The same deficiency leads liberals
to read only The Nation and conservatives to read only National Review. We
are not natural falsificationists: we would rather find more reasons for
believing what we already believe than look for reasons that we might be
wrong. In the terms of Karl Popper's famous example, to verify our intuition
that all swans are white we look for lots more white swans, when what we
should really be looking for is one black swan.
Also, people tend to see the future as indeterminate and the past as
inevitable. If you look backward, the dots that lead up to Hitler or the
fall of the Soviet Union or the attacks on September 11th all connect. If
you look forward, it's just a random scatter of dots, many potential chains
of causation leading to many possible outcomes. We have no idea today how
tomorrow's invasion of a foreign land is going to go; after the invasion, we
can actually persuade ourselves that we knew all along. The result seems
inevitable, and therefore predictable. Tetlock found that, consistent with
this asymmetry, experts routinely misremembered the degree of probability
they had assigned to an event after it came to pass. They claimed to have
predicted what happened with a higher degree of certainty than, according to
the record, they really did. When this was pointed out to them, by Tetlock's
researchers, they sometimes became defensive.
And, like most of us, experts violate a fundamental rule of probabilities by
tending to find scenarios with more variables more likely. If a prediction
needs two independent things to happen in order for it to be true, its
probability is the product of the probability of each of the things it
depends on. If there is a one-in-three chance of x and a one-in-four chance
of y, the probability of both x and y occurring is one in twelve. But we
often feel instinctively that if the two events "fit together" in some
scenario the chance of both is greater, not less. The classic "Linda
problem" is an analogous case. In this experiment, subjects are told, "Linda
is thirty-one years old, single, outspoken, and very bright. She majored in
philosophy. As a student, she was deeply concerned with issues of
discrimination and social justice and also participated in antinuclear
demonstrations." They are then asked to rank the probability of several
possible descriptions of Linda today. Two of them are "bank teller" and
"bank teller and active in the feminist movement." People rank the second
description higher than the first, even though, logically, its likelihood is
smaller, because it requires two things to be true-that Linda is a bank
teller and that Linda is an active feminist-rather than one.
Plausible detail makes us believers. When subjects were given a choice
between an insurance policy that covered hospitalization for any reason and
a policy that covered hospitalization for all accidents and diseases, they
were willing to pay a higher premium for the second policy, because the
added detail gave them a more vivid picture of the circumstances in which it
might be needed. In 1982, an experiment was done with professional
forecasters and planners. One group was asked to assess the probability of
"a complete suspension of diplomatic relations between the U.S. and the
Soviet Union, sometime in 1983," and another group was asked to assess the
probability of "a Russian invasion of Poland, and a complete suspension of
diplomatic relations between the U.S. and the Soviet Union, sometime in
1983." The experts judged the second scenario more likely than the first,
even though it required two separate events to occur. They were seduced by
the detail.
It was no news to Tetlock, therefore, that experts got beaten by formulas.
But he does believe that he discovered something about why some people make
better forecasters than other people. It has to do not with what the experts
believe but with the way they think. Tetlock uses Isaiah Berlin's metaphor
from Archilochus, from his essay on Tolstoy, "The Hedgehog and the Fox," to
illustrate the difference. He says:
Low scorers look like hedgehogs: thinkers who "know one big thing,"
aggressively extend the explanatory reach of that one big thing into new
domains, display bristly impatience with those who "do not get it," and
express considerable confidence that they are already pretty proficient
forecasters, at least in the long term. High scorers look like foxes:
thinkers who know many small things (tricks of their trade), are skeptical
of grand schemes, see explanation and prediction not as deductive exercises
but rather as exercises in flexible "ad hocery" that require stitching
together diverse sources of information, and are rather diffident about
their own forecasting prowess.
A hedgehog is a person who sees international affairs to be ultimately
determined by a single bottom-line force: balance-of-power considerations,
or the clash of civilizations, or globalization and the spread of free
markets. A hedgehog is the kind of person who holds a great-man theory of
history, according to which the Cold War does not end if there is no Ronald
Reagan. Or he or she might adhere to the "actor-dispensability thesis,"
according to which Soviet Communism was doomed no matter what. Whatever it
is, the big idea, and that idea alone, dictates the probable outcome of
events. For the hedgehog, therefore, predictions that fail are only "off on
timing," or are "almost right," derailed by an unforeseeable accident. There
are always little swerves in the short run, but the long run irons them out.
Foxes, on the other hand, don't see a single determining explanation in
history. They tend, Tetlock says, "to see the world as a shifting mixture of
self-fulfilling and self-negating prophecies: self-fulfilling ones in which
success breeds success, and failure, failure but only up to a point, and
then self-negating prophecies kick in as people recognize that things have
gone too far."
Tetlock did not find, in his sample, any significant correlation between how
experts think and what their politics are. His hedgehogs were liberal as
well as conservative, and the same with his foxes. (Hedgehogs were, of
course, more likely to be extreme politically, whether rightist or leftist.)
He also did not find that his foxes scored higher because they were more
cautious-that their appreciation of complexity made them less likely to
offer firm predictions. Unlike hedgehogs, who actually performed worse in
areas in which they specialized, foxes enjoyed a modest benefit from
expertise. Hedgehogs routinely over-predicted: twenty per cent of the
outcomes that hedgehogs claimed were impossible or nearly impossible came to
pass, versus ten per cent for the foxes. More than thirty per cent of the
outcomes that hedgehogs thought were sure or near-sure did not, against
twenty per cent for foxes.
The upside of being a hedgehog, though, is that when you're right you can be
really and spectacularly right. Great scientists, for example, are often
hedgehogs. They value parsimony, the simpler solution over the more complex.
In world affairs, parsimony may be a liability-but, even there, there can be
traps in the kind of highly integrative thinking that is characteristic of
foxes. Elsewhere, Tetlock has published an analysis of the political
reasoning of Winston Churchill. Churchill was not a man who let
contradictory information interfere with his idées fixes. This led him to
make the wrong prediction about Indian independence, which he opposed. But
it led him to be right about Hitler. He was never distracted by the
contingencies that might combine to make the elimination of Hitler
unnecessary.
Tetlock also has an unscientific point to make, which is that "we as a
society would be better off if participants in policy debates stated their
beliefs in testable forms"-that is, as probabilities-"monitored their
forecasting performance, and honored their reputational bets." He thinks
that we're suffering from our primitive attraction to deterministic,
overconfident hedgehogs. It's true that the only thing the electronic media
like better than a hedgehog is two hedgehogs who don't agree. Tetlock notes,
sadly, a point that Richard Posner has made about these kinds of public
intellectuals, which is that most of them are dealing in "solidarity" goods,
not "credence" goods. Their analyses and predictions are tailored to make
their ideological brethren feel good-more white swans for the white-swan
camp. A prediction, in this context, is just an exclamation point added to
an analysis. Liberals want to hear that whatever conservatives are up to is
bound to go badly; when the argument gets more nuanced, they change the
channel. On radio and television and the editorial page, the line between
expertise and advocacy is very blurry, and pundits behave exactly the way
Tetlock says they will. Bush Administration loyalists say that their
predictions about postwar Iraq were correct, just a little off on timing;
pro-invasion liberals who are now trying to dissociate themselves from an
adventure gone bad insist that though they may have sounded a false alarm,
they erred "in the right direction"-not really a mistake at all.
The same blurring characterizes professional forecasters as well. The
predictions on cable news commentary shows do not have life-and-death side
effects, but the predictions of people in the C.I.A. and the Pentagon
plainly do. It's possible that the psychologists have something to teach
those people, and, no doubt, psychologists are consulted. Still, the
suggestion that we can improve expert judgment by applying the lessons of
cognitive science and probability theory belongs to the abiding modern
American faith in expertise. As a professional, Tetlock is, after all, an
expert, and he would like to believe in expertise. So he is distressed that
political forecasters turn out to be as unreliable as the psychological
literature predicted, but heartened to think that there might be a way of
raising the standard. The hope for a little more accountability is hard to
dissent from. It would be nice if there were fewer partisans on television
disguised as "analysts" and "experts" (and who would not want to see more
foxes?). But the best lesson of Tetlock's book may be the one that he seems
most reluctant to draw: Think for yourself.
source: www.newyorker.com/critics/books/articles/051205crbo_books1