Previous Month | RSS/XML | Current | Next Month

WEBLOG

June 30th, 2017 (Permalink)

The Limits of Experts

I've been skeptical of Tom Nichols' new book The Death of Expertise1, but perhaps that's because it reminds me of former Book Club book Wrong2, which lived up to its title. Now, however, there is another excerpt from the book, entitled "The Crisis of Expertise"3, which is worth a look.

I'm just as skeptical that there is a "crisis" of expertise as I am about it's alleged death2. "Crisis", like "death", is one of those scare words intended to draw attention and motivate to action: in this case, to at least read a magazine article. I recommend that you do so, but not because there's a "crisis".

There is, however, a long-standing problem of expertise, namely, how a layperson can tell a genuine expert from a pseudo-expert. This problem is exacerbated by the fact that often the same person is both, that is, an expert about one thing but a pseudo-expert about another. Here's Nichols on that sub-problem:

One of the most common errors that experts make is to assume that, because they are smarter than most people about certain things, they are smarter than everyone about everything. Overconfidence leads experts not only to get out of their own lane and make pronouncements on matters far afield of their expertise, but also to ‘over-claim’ wider expertise even within their own general area of competence. Experts and professionals, just as people in other endeavours, assume that their previous successes and achievements are evidence of their superior knowledge, and they push their boundaries rather than say the three words that every expert hates to say: ‘I don’t know.’ No one wants to appear to be uninformed or to be caught out on some ellipsis in their personal knowledge. Laypeople and experts alike will issue confident statements on things about which they know nothing, but experts are supposed to know better. … The public is remarkably tolerant of such trespasses, and this itself is a paradox: while some laypeople do not respect an expert’s actual area of knowledge, others assume that expertise and achievement are so generic that experts and intellectuals can weigh in with some authority on almost anything.3

Nichols proceeds to give some prominent examples of experts making fools of themselves outside of their areas of expertise. While there may be no "crisis" of expertise, this is a perennial problem, especially when it comes to politics. The fact that this is not a recent phenomenon is shown by Nichols' examples: Linus Pauling died over twenty years ago4 and the example involving Helen Caldicott is from 1983. Whatever truth there may be to the claim that the problem of expertise is worsening, I don't think it's because experts are worse now than they've ever been about over-reaching. Back to Nichols:

Day to day, laypeople have no choice but to trust experts. … This is not the same thing as trusting professionals when it comes to matters of public policy: to say that we trust our doctors to write us the correct prescription is not the same thing as saying that we trust all medical professionals about whether the US should have a system of national healthcare. To say that we trust a college professor to teach our sons and daughters the history of the Second World War is not the same as saying that we therefore trust all academic historians to advise the president of the US on matters of war and peace. For these larger decisions, there are no licences or certificates. There are no fines or suspensions if things go wrong. Indeed, there is very little direct accountability at all, which is why laypeople understandably fear the influence of experts.3

There are no experts about "whether the US should have a system of national healthcare", and other questions of what political policies we should adopt. Of course, there are experts in the sense that they know much more than you or I do about Obamacare, or what's in the healthcare plan currently wending its way through congress, and we should consult with them to learn such things. However, what type of healthcare system we should adopt is a political decision ultimately depending on ethical judgments, and there are no experts who can tell us what judgments to make.

One thing that Nichols writes that I think is not quite right is the following passage:

Prediction is a problem for experts. It’s what the public wants, but experts usually aren’t very good at it. This is because they’re not supposed to be good at it; the purpose of science is to explain, not to predict. And yet predictions, like cross-expertise transgressions, are catnip to experts. These predictions are often startlingly bad.3

Prediction is a part of the scientific method because theories are often tested by using them to make predictions, and then checking whether the predicted event occurs. I suspect that Nichols is overstating his case because he's thinking in terms of his own area of expertise, which is political science, and related areas of social science such as economics and sociology. How well a science can predict things depends upon whether it is a natural or social science. For instance, astronomers can make very accurate predictions of the positions of planets in the sky, the return of comets, and the occurrence of eclipses.

However, sciences that deal with human beings are different, partly because people are unpredictable, but there is also an inherent problem that such predictions may affect whether the predicted event occurs. For instance, a prominent economist's prediction of a recession may contribute to a loss of confidence in the economy that triggers a recession, or a prediction of an economic boom may lead to increased confidence that prevents a recession5. For this reason, despite the overconfident predictions of some 19th-century philosophers and scientists6, there can be no social science able to predict the course of human history.

I now have access to a copy of the book, so I may have something more to say about it later this year.

Notes:

  1. New Book: The Death of Expertise, 2/28/2017
  2. Wrong, Again, 3/24/2017
  3. Tom Nichols, "The Crisis of Expertise", Aeon, 6/6/2017
  4. "Linus Pauling―Biographical", Nobel Prize, 2014
  5. This is over-simplified, of course, in that the pronouncements of no single economist would probably have so much of an effect on the economy, but what people think about the economy does affect it, and what economists predict can affect what people think.
  6. Comte, Hegel, Marx, to name a few.

Update (7/1/2017): Friend of The Fallacy Files Patricia Heil emailed a good point about Nichols' claim, quoted above, that experts are "smarter" in their areas of expertise than other people:

It's clear the text has problems like the title. Experts are not smarter than other people. They have more training. This is a chronic problem in the US if not other places. Intelligence or being smart is a capability. Expertise is the result of knowledge and experience, that is, education. I tell people: you take Einstein and raise him in the Russian steppes instead of Germany, and he may still be intelligent enough to learn math easily, but he's not going to describe relativity because he doesn't get an education. The people like Linus Pauling who sound off in fields where they've never trained or worked are victims of this problem.
June 11th, 2017 (Permalink)

Cuckoo for Cocoa Puffery

There are three sets of words to be on the lookout for in media reports of statistical studies of medicine or nutrition:

  1. "Could", "may", "possible", "suggest", and other words indicating that the results of a study are weak.
  2. "Link", "association", "connection", "relationship", and other words that mean that two variables are related statistically, but not necessarily causally.
  3. Other words, such as "help", that indicate that any possible effect is probably small.

With that in mind, I have highlighted every occurrence of these words in the following excerpt from a recent news story1. They first make an appearance in the headline:

Eating chocolate may help prevent a fairly common heart problem

Eating chocolate has been linked with a reduced risk of heart disease and stroke. Now a study from Denmark suggests that regular consumption of the treat may help to prevent the development of atrial fibrillation, a type of irregular heartbeat.

In the study…researchers found that people who ate chocolate at least once a month had rates of atrial fibrillation that were 10 to 20 percent lower than those who ate chocolate less often. …

When researchers took into consideration other factors that might influence development of atrial fibrillation, such as alcohol intake, smoking, obesity, high blood pressure and cholesterol levels, the study showed an association between people with a moderate intake of chocolate and a lower risk of developing atrial fibrillation.

The study does not prove a cause-and-effect relationship. And although the exact mechanism of how chocolate may prevent atrial fibrillation is not known, it’s possible that compounds in chocolate called flavonoids may play a role, the researchers said.

Flavonoids have antioxidant and anti-inflammatory properties…. They may limit the inflammatory process in the body, reducing the stickiness of the blood and leading to less scarring of connective tissue. All of these factors may help prevent the electrical remodeling of the heart that leads to atrial fibrillation….

The findings showed that for women, the strongest association was seen in those who ate a one-ounce serving of chocolate once a week: This level of consumption was linked to a 21 percent lower risk of atrial fibrillation compared with those who ate less chocolate. For men, the strongest association was seen in those who ate two to six one-ounce servings of chocolate weekly. These men had a 23 percent lower risk of atrial fibrillation. …

All in all, the findings suggest that compared with some other snack choices, a moderate intake of chocolate may be a heart-healthy snack….

If only The Washington Post and other news media would publish health articles with these three types of word highlighted, as above, then it would be obvious how weak the studies often are. I don't pick on the above story on the grounds that it is especially bad; rather, it's a typical example of this type of reporting.

It's also typical in its first-the-good-news-then-the-bad-news structure: in the first half, we're given the good news: chocolate is good for you! In the second half, we're told the bad news: never mind! This is why, as I've noted before2, you should read these articles all the way to the end. Unless, of course, all you want to read is the fake news.

Notes:

  1. Cari Nierenberg, "Eating chocolate may help prevent a fairly common heart problem", The Washington Post, 5/27/2017
  2. Caveat Lector, 9/3/2011
June 9th, 2017 (Permalink)

Puzzle: Your Mileage May Differ

A family owns two cars: a sport utility vehicle that gets 20 miles per gallon and a compact car that gets 40 MPG. Both vehicles are getting old and the family has decided to trade one in on a replacement that gets better mileage on a tank of gasoline. The possibilities are switching the SUV for one that gets 25 MPG or replacing the compact with one that gets 50 MPG. They drive each car about the same amount. Assuming that the only relevant consideration is how much they can save on gas, which vehicle should they trade in?

Solution

June 8th, 2017 (Permalink)

Amphibolous Headline

Teen accused of killing Uber driver with machete in one hand, knife in the other1

My first thought when I saw this headline was: if an Uber driver came at me with a machete in one hand and a knife in the other, I guess I'd kill him, too. However, if you read the story beneath the headline, it was the teen who is accused of using a machete and knife to kill the driver.

This headline isn't funny and neither is the story beneath it. However, it is ripe for misinterpretation since the headline is ambiguous; more specifically, it is amphibolous2 because of a misplaced modifier3. Grammatically, the modifying phrase "with machete in one hand, knife in the other" should modify "Uber driver", which it immediately follows, and not "teen" or "killing". Using brackets to indicate the scope of the modifying phrase, three distinct meanings of the headline can be disambiguated:

  1. Teen accused of killing [Uber driver with machete in one hand, knife in the other]

    The phrase acts as an adjective, modifying "Uber driver", which means that the driver had the weapons.

  2. [Teen with machete in one hand, knife in the other] accused of killing Uber driver

    Again, the phrase acts as an adjective, but now modifies "teen". The teen had the weapons, but when accused, not necessarily while killing.

  3. Teen accused of [killing, with machete in one hand, knife in the other] Uber driver

    The phrase now acts as an adverb, modifying the verb "killing" rather than a noun. This version accurately represents what the news story reported the teen was accused of doing.

I can sympathise with the editor who wrote the headline, since it's difficult to pack so much information into a brief headline unambiguously. However, here's a suggested revision:

Teen accused of using a machete in one hand and a knife in the other to kill Uber driver

This is four words longer but three times clearer.

Notes:

  1. Samantha Schmidt, "Teen accused of killing Uber driver with machete in one hand, knife in the other", The Washington Post, 6/1/2017
  2. See Amphiboly.
  3. See Robert J. Gula, Precision: A Reference Handbook for Writers (1980), sections 10B & 10C

Solution to Your Mileage May Differ: Surprisingly enough, the family should replace the SUV. It looks as though they ought to trade in the compact car, since the replacement will get 10 MPG better mileage, whereas the new SUV will only get 5 MPG more. Since they drive the two cars about the same distance, it seems as though they would save more gas with the new compact car rather than the SUV.

Here's why they should select the SUV: suppose that they drive each vehicle 10,000 miles a year. Then, since the SUV gets 20 MPG, it uses 500 gallons of gas a year, and because the compact car gets 40 MPG, it uses half as much gas, namely, 250 gallons per year. An SUV that gets 25 MPG will use only 400 gallons a year, thus saving the family 100 gallons; whereas a compact car that gets 50 MPG, will use 200 gallons of gas in a year, for a savings of 50 gallons. They can save twice as much gas by trading in the SUV!

Why is this counter-intuitive result correct? Clearly, if both cars got the same MPG to start with, then the one to replace would be the one that most improves that MPG. However, there is a big difference in the MPGs of the two cars, with the compact getting twice the mileage on a tank of gas that the SUV gets.

Instead of looking at the absolute difference in MPGs, consider the percentage improvement: from 20 to 25 is a 25% improvement, and from 40 to 50 is also a 25% improvement. So, they're the same percentage improvement! However, because the SUV uses twice as much gas as the compact, a 25% improvement in mileage will actually save twice as much gas.

Source: Derrick Niederman & David Boyum, What the Numbers Say: A Field Guide to Mastering Our Numerical World (2003), pp. 65-66.

Previous Month | RSS/XML | Current | Next Month