Being the guy who wrote a book called Ultralearning [1], I get asked a lot of questions about what I think schools should be doing better. Having never taught in a classroom or worked for even a single day in education, it’s a question I’m totally unqualified to answer. It’s a bit like asking a guy to reform an entire health care system because he’s good at lifting weights.
But being totally unqualified has never stopped me before, so I’ll try to explain the answer I typically give to this question, which is that I’m skeptical of dramatic proposals to make school considerably more effective or efficient for the average student.
To be clear, that’s not because no improvement is possible. We do know some about things that work that are inconsistently applied: phonics [2] should be taught, cognitive load [3] should be managed, skills should be fully taught [4] and practice [5] should be fun and ample.
But these answers aren’t the kind that satisfy the people who ask me these questions. Instead, having had many of these conversations, I feel like the person asking already “knows” what my response should be:
Isn’t it obvious that school sucks? That we should be teaching critical thinking and problem-solving skills instead of useless facts and theories? That school should be more like real life, with real-world projects and experiments and collaboration? That there should be less of that stuffy work of sitting in a desk and memorizing things?
If you had asked me this question years ago, I probably would have agreed with you. It took reading a lot of research to convince me that this intuitively appealing idea is actually bad. Below, I’d like to explain why.
First, the Evidence
Before I get into the explanation of why these kinds of seemingly-good strategies don’t work, I should begin by pointing out that these ideas are not new. They have been tried, and they have been found wanting.
Entire books have been written pointing out the flaws in many of these strategies. I won’t be able to do the full debate justice here, but, if you’re interested, you can check out Daniel Willingham’s Why Don’t Students Like School? [6] Greg Ashman’s The Power of Explicit Teaching and Direct Instruction [7] or, if you want to learn more about the actual debate between proponents of both sides, try Constructivist Instruction: Success or Failure [8].
To briefly recap some of the evidence:
- Project Follow Through [9] was one of the largest educational experiments ever conducted. Run in the 1970s, it compared how different teaching methodologies impact student outcomes. Direct Instruction [10], a method of teaching that has students sit in desks and perform extremely structured drills in unison, performed best.
- Problem-based learning [11] tends to do worse than traditional schooling in medical education. An influential meta-analysis [12] by Albanese and Mitchell, for instance, found that students required more time studying, had worse exam scores and ordered more unnecessary tests compared to traditionally taught students.
- Despite needing to relearn this truth every few decades, the best way to teach kids how to read has been known for centuries [2]: break down the sound-spelling correspondence, and do lots of practice on it before moving up to authentic texts. Approaches based on skipping these drills in favor of “inspiring a love of reading” do worse.
- Practice testing and distributed practice—basically, having regular quizzes spread out over a course—are the studying methods with the best empirical support [13]. Fancy methods like mnemonics and concept maps fare worse.
- General problem solving abilities are neither learned nor taught [14]. While some problem-solving methods have broader applicability than others (such as the scientific process of hypothesis testing), students learn these methods better [15] when they’re explicitly taught rather than simply giving students projects and hoping they’ll reinvent them on their own.
In short, whenever we have high-quality evidence that rigorously compares two teaching methods, the research invariably favors strong, direct instruction plus practice.1 [16] Or, in other words, the exact stereotype of schooling that so many of the people asking me about school reform despise.
Your Stereotype of School is an Endangered Species
This doesn’t mean education couldn’t be better. My impression upon first encountering the Direct Instruction research was that I had never been taught this way in my entire life.
Clichés are often out of date. I went to grade school in the nineties, when the lofty aims of project-based and discovery learning were the educational orthodoxy. I spent a lot of my school years doing time-consuming projects that had us gluing and coloring and expected us to do our own research.
I’m pessimistic about real reform because the changes needed to make schools more effective are often opposite of what many people intuitively feel. For schools to teach more effectively, they should be more rigorous about carefully defining the knowledge objectives of the class, thoroughly breaking down complex skills into components, and doing lots and lots and lots of practice.
In short, a “better” school probably looks more like the stereotype of an old-fashioned schoolhouse with kids sitting in desks, drilling facts and concepts that are patiently explained by a teacher. To the extent that school becomes more like free play, project-building or acting like a scientist, it will probably be worse.
Why It’s Hard to Improve Schools
Schools face a number of practical constraints that make thorough reform difficult. Students are unmotivated. They range in background knowledge and innate ability. We care about sorting just as much as educating, so schools end up doing both.
But the real reason it’s hard to improve schools is simply that there are fundamental constraints on how the brain learns that prevent radical shortcuts.
The boring truth is that expertise in most subjects is largely a matter of having an enormous library of knowledge and skill. For example, if you want to learn a language, you need to learn a lot of words. Any method that tries to skip over the fact that there are tens of thousands of words to learn is doomed to failure. All skills are like this, it’s simply that the “atoms” of learning are usually less obvious than in languages.
When students complain about all the boring facts and skills they had to learn in school, my response is to claim that there isn’t any other type! All skills are simply an accumulation of small bits of facts, procedures and concepts.
Those small bits, in isolation, seem kind of trivial. But quantity has a quality all its own, and with enough well-integrated knowledge the result is expertise that seems almost magical to those who don’t possess it.
This means that improving education comes down to largely two different options:
First, you can increase the efficiency of the system. Efficiency here looks like the kind of factory redesign that increases product throughput—increasing the number of words learned per day, optimizing cognitive load, boosting mnemonic efficiency through spacing and retrieval—without skipping over the fundamental bottleneck in cognition.
Second, you can choose to learn different things. Given the high degree of specificity of most knowledge, the choice of what to learn can have profound consequences. But if choosing a pedagogical method is contentious, curricular choice is even more so! For every “useless” subject that reformers want to discard, there are die-hard advocates arguing that we should be putting in even more.
I believe in both of these things, and I’ve focused much of my writing career on how we can do them better, particularly outside of the typical classroom. But if you want lots of skills, there’s no way around learning a lot of stuff—including a ton of stuff that feels too obscure to be broadly useful.
What About Ed Tech?
Thus far, I’ve mostly been targeting a certain kind of questioner: someone who feels that school was maybe too boring and impractical, and who longs for the possibility that education could be more like play and less like studying.
There are reformers of all stripes, and educational technologists are another side of this debate. These are the people who champion efforts to gamify learning, carefully match teaching to each student’s ability level, develop AI-based tutoring, put an iPad in every child’s hands and so on.
In theory, these ideas are possibly useful. Drills can be boring, so wrapping them with gamification elements that reward progress and engagement might be helpful. Skills can be too hard or too easy [17], so adjusting difficulty automatically might be helpful. AI-tutoring, too, might help with closing Bloom’s famous 2 sigma problem [18].
But I’m more skeptical in practice. As Kelsey Piper writes [19], a lot of ed-tech games have a fairly low density of actual useful learning. I can attest to this: eager to give my son a head start on the phonetic skills involved in reading, I tried a few different iPad games with him. He mostly messed around randomly until he got the reward, largely ignoring the educational content to fixate on the cute cartoon characters.
Gamified learning is a bit like wrapping medicine in candy. Yes, it may help some students swallow some instruction they otherwise find bitter, but in practice it’s easy to pull off the candy, consume it, and throw the medicine away.
Individualized instruction aided by technology does solve some of the problems of differing ability levels. But schools aren’t just solving the cognitive problems of learning, they’re also working on motivational ones. A rigorous, but achievable, standard that applies to everyone may be more sustainable for motivation than an individually-tailored goal for each student.2 [20]
Similarly, while I’m hopeful that AI advances will make automated tutoring more useful, it’s still far away from the skill a teacher can provide. As someone who makes use of AI [21] quite a bit in my own learning, I can say that it’s still relatively weak at having a good model of an individual’s skill gaps and conceptual weaknesses. It’s very much at the “better than nothing”—not the “better than teachers”—stage right now.
So, while I’m hopeful that there will be some improvements in technology around the margins, I’m skeptical of anything touted as a radical overhaul in educational process or outcomes.
What About Ultralearning?
Does this line of thinking rule out the methods I describe in Ultralearning [1]? I don’t think so. The big distinction (and it is a big one) between my aims for that book and the aims of educational reformers is that I started with the assumption of a highly motivated learner.
The people I document in that book all began with the starting point that they were willing to work hard, even obsessively, on a project they were deeply motivated to succeed with. In such cases, the classroom structures that facilitate motivation can instead become obstacles: fixed homework assignments, mandatory lectures, exam deadlines. These things keep uninterested students going, but they may hold back the aggressively curious.
For instance, I still believe that full immersion [22] is the best way to learn a language, provided you also augment that with a lot of the studying approaches I describe above, but it’s obviously a high-effort strategy. I’ve spoken to lots of people who have asked me for advice on how to learn a language, but relatively few that took up Vat’s and my actual strategy of avoiding speaking English [23].
Ultimately, I believe enhanced learning is certainly possible, and a highly motivated person can often do better than the average, and sometimes even the upper end, of what is typically seen in school. But such optimism about the possibility of learning doesn’t so easily transfer to a situation where motivation is much lower.
Footnotes
- Manu Kapur’s work on productive failure [24] doesn’t undermine this finding, contrary to some misinterpretations. Kapur’s research simply finds that the timing of instruction has some effects. Sometimes, for certain kinds of skills, in certain kinds of environments, attempting to solve a problem first and failing can be helpful for later understanding the solution procedure that is fully taught.
- Researcher Greg Ashman makes a good argument against common pleas to individualize instruction in his book [25].