5 min read

Be Wrong First

learning sciencememorymetacognition

Guessing at a question before you've studied it, and getting it completely wrong, makes you learn the right answer better than if you'd spent those same minutes reading. I didn't believe it the first time I came across it either, so this is about why it holds up, and why almost every study tool is built in a way that stops it happening.

Everyone runs the same order. Learn the thing, then test yourself on it, because testing is for checking and studying is for learning, and you wouldn't sit an exam on a chapter you hadn't opened yet.

The research says flip it. Attempt the questions first, fail them, then study, and you come out ahead of someone who spent that entire time studying instead. It's called the pretesting effect (Richland, Kornell and Kao, 2009), and the failing isn't a tax you're paying for the benefit, the failing is where the benefit comes from (Kornell, Hays and Bjork, 2009).

What's actually happening when you guess wrong

Say you're about to start a chapter on inflation and you haven't read a word of it yet. Someone asks you why printing more money pushes prices up. You don't know, but you're not blank either, you've heard something somewhere about more money chasing the same goods, so you commit to that, and you get told you've got roughly the right shape but you're missing why the goods don't just increase to match.

Thirty seconds, and two useful things happened in them.

The first is that you dragged everything you vaguely already knew about money out into the open. That matters, because the correct explanation now has somewhere to attach itself instead of landing on blank ground, and having somewhere to attach is most of what makes a thing memorable in the first place (Kornell, Hays and Bjork, 2009).

The second is that you now genuinely want to know. You made a claim, you were told it was incomplete, and there's an open question sitting in your head that wasn't there a minute ago. So when the explanation turns up it arrives as the answer to something you were already holding rather than as one more sentence to get through, and information that closes a question you were carrying sticks far better than the same information handed to you cold (Slamecka and Graf, 1978).

Neither of those happens if you just read the chapter.

And you get a map out of it for free

There's a second thing a pre-test does that gets mentioned far less, which is that it shows you where your gaps are before you've sunk three hours into the chapter.

Most revision is undirected. You reread the whole thing because you honestly can't tell which parts you have and which you don't, and there's a trap hiding in that: the parts you already know are the nicest bits to read. They go quickly, they feel like progress, nothing catches. So you quietly end up spending the most time on the material you needed the least. Ten minutes of guessing turns "revise chapter 7" into "revise these four things", and those four things are where the time should have been going from the start.

So why is nearly every tool built to punish it?

If getting things wrong is the mechanism, then marking it in red and playing a happy sound for the green ones is precisely the wrong instinct. It's the default almost everywhere though, and I think that's mostly because right-and-wrong is trivial to build and it looks like feedback.

The cost only becomes obvious when you think about who it lands on. A perfectionist, an overachiever, or a lot of neurodivergent learners will read a red cross as a failure, and the anxiety that follows is exactly the thing a pre-test needs switched off in order to work. The entire mechanism depends on you being willing to write down an answer you're fairly sure is wrong. Make that feel bad and people stop committing to answers. They hedge, or they skip the pre-test, and the moment the guessing goes the effect goes with it.

The design rule

If being wrong is the point of a feature, the interface isn't allowed to punish it. No red, no score, no streak to break. Just here's what you said, here's what's true, carry on.

That's why there's no red cross anywhere in Norudit's pre-test. You answer everything, then you see your own answer sitting next to the real one, with any misconception you're holding written out plainly rather than scored. It's the single part of the app where getting things wrong is the intended outcome, so designing it like something you could fail would have been a strange decision.

There's a small amount of courage involved in writing down an answer you already suspect is wrong. The least a study tool can do is not charge you for it.

Sources

Read more