Annoying, Boring, Correct: The ABCs of good social science
If it is surprising and satisfying, chances are good it's spurious.
The social sciences have been thoroughly unreliable for the last half-century, as demonstrated repeatedly in the last decade of the replication crisis. This is not a matter of a simple failure of any one theoretical model. It is not because social scientists don’t know statistics or epistemology. Frankly, psychologists are remarkably well-informed about the philosophy of science.
This is a problem with roots in the peer review system, publication incentives, and a huge public appetite for social science results that are surprising and satisfying, even if those results are spurious. Correctly done social science is often annoying or boring instead, and this means that false results can travel halfway around the world before the truth has finished tying its shoelaces.
Surprising, satisfying, and spurious
People love social science results that are surprising and satisfying. In 2014, a paper was published in the Proceedings of the National Academy of Sciences (PNAS) claiming that hurricanes with feminine names caused more damage because people didn’t take them seriously. This was surprising. It was also satisfying for some people on ideological grounds, the result being that the study has continued to be breathlessly cited by reporters years after it was debunked as totally spurious.
Conservatives are just as susceptible as feminists to the lure of surprising and satisfying ideas dressed up as serious social science. David Stockman wrote that in January of 1980, when presidential contender Ronald Reagan was introduced to the idea that one could increase government revenue by cutting taxes, it “set off a symphony in his ears” and he “knew instantly it was true.” Reagan wanted to believe that he could cut taxes and address the budget deficit at the same time. Instead, his tax cuts blew a hole in the national deficit.1
Not all defective social science is explicitly ideological, although ideology plays an important role. Educators wanted to believe that “whole language” reading instruction would work better than boring old phonics, setting back literacy rates across the United States.2 At the end of the day, everyone seems to want to find the One Weird Trick that will make teaching math fun and easy.3
Annoying, boring, correct
Direct instruction works. Professors have been lecturing students systematically for as long as professors have existed. Classrooms full of desks with students listening, reading, and taking notes have educated generations and generations of students. It’s boring. It’s annoying. It’s correct. Many teachers want to break the mold, but acting like Robin Williams in Dead Poets Society usually just gets you bad student evals, not students standing on desks in solidarity.
Outside of truly extraordinary circumstances, budget deficits are most easily addressed by raising taxes, cutting spending, or both. Making people use checklists reduces error rates. Most men are attracted to women with slender waistlines and most women prefer taller men. If prices rise, demand will usually fall in response.
One of the few solid new developments in the last several decades of psychology: Human memory is surprisingly bad. It’s inaccurate, fuzzy, and malleable. Memories change over time. Eyewitnesses identify the wrong person all the time. False memories can be fabricated from whole cloth. While this is surprising, it is also extraordinarily annoying; in fact, it is annoying enough to be broadly controversial, especially among therapists who believe in searching for repressed memories.4
One example I talked about recently: The median voter theorem says that in general, given some particular assumptions, more moderate candidates will perform better. This top-level message is boringly obvious. While it’s not annoying in general, it’s annoying to many reliable partisans … which includes most working political scientists in the United States and most of the highly politically engaged people in the United States.
It’s much more exciting to claim that highly partisan candidates with extreme and polarizing messages will do better. Strongly partisan Democratic political scientists have gotten a lot of air time with strongly partisan Democratic media figures by embracing the surprising, satisfying, and largely unsupported claim that Democrats needed to lean hard into partisanship and nominate a radical candidate who would generate enthusiasm among minority and left-wing voters.5
It was in that media environment that I wrote my “blue bubble” article arguing that Democrats should nominate a moderate if they wanted the best chance of winning. Democrats proceeded to nominate one of the most moderate available candidates (Biden). In a show of boring predictability, a moderate presidential candidate significantly outperformed most of the other Democrats on the ballot.
Make social science boring again
The consequence of rewarding surprising and satisfying studies in the social sciences has been an astonishing rise in misinformation in the social sciences. A psychology student who graduated in 2010 was confidently told a larger number of demonstrably incorrect “facts” about human psychology than a psychology student who graduated in 1960. By some measures, the social sciences are going backwards.
I say 2010, because that’s about when I was a teaching assistant in an introductory psych class. A clear majority of the psychology experiments discussed in that class that took place between 1960 and 2010 have either failed to replicate or been investigated and shown fraudulent. That’s not to say that there aren’t people steadily plugging away at advancing the social sciences, but the boring parts rarely make the news and barely get noticed by scientific journals and tenure review committees, which is a problem for all of us.
If a study sounds surprising and satisfying to you, be skeptical, for it may be spurious.
Subsequent serious work by macroeconomists has suggested that the peak of the Laffer curve tends to be somewhere around 70%, with only a handful of countries exceeding their peak revenue-generating tax rate at any point between 1980 and the present.
That is, teaching whole words as if they were logograms, rather than teaching children the sounds associated with letters of a phonetic alphabet and having them sound out unfamilar words phonetically. A line from one meta-analysis, published in 2018, drily puts it this way: “Our experience is that once the nature of the writing system is understood, the importance of phonics instruction in the initial stages of learning to read becomes obvious.”
When I went to the Joint Math Meetings for the first time, the “flipped classroom” was the new sexy trick that would solve everything. Promising pilot studies, lots of people very excited… then it turned out students really hated it. Maybe I’ll tell my story about that sometime.
Yet another episode in the long list of ways in which Freudian psychology is harmful, which has been directly implicated in the “satanic panic” of the 1980s as well as other luridly false but sincere allegations.
The fact that these were treated as synonymous by many commentators is not grounded in any kind of empirical science.