Ease of use is not ease of learning

The situation

Commercial product design has a clear target. It wants more use: more sessions, longer sessions, more done in each one, and fewer people leaving. Nearly every method in the field serves that target. Reduce friction. Shorten the path to the thing the user wants. Personalise toward what they already like. Reward the action you want repeated. These work, and they work because the product’s success and the user’s satisfaction are close to the same quantity. A person who enjoys a shopping app buys more, and the app is doing its job.

Learning products are mostly built by people trained on that target. Designers, developers and product owners arrive from consumer software, and their instincts arrive with them. So the learning product gets a smooth onboarding, a streak counter, a recommendation engine that serves more of what the learner is already good at, and a review process that treats every moment of difficulty as a defect to be filed. The engagement figures look fine. They are the wrong figures.

I have sat in those reviews. I have also been the person asking for the friction to go, on the grounds that the learner seemed to be struggling. They were. That was the point.

The question

What if the quantity being optimised is the wrong one? In a learning product the user’s satisfaction and the product’s success come apart. The experience that feels best now is often the one that teaches least, and the experience that teaches most often feels like failing while it happens.

That is a finding, not a hunch. Roediger and Karpicke (2006) had students either re-read a passage or test themselves on it. Five minutes later the re-readers did better. A week later the testers did, by a wide margin. Kornell and Bjork (2008) showed learners that spacing beat massing in their own results, and the learners still rated massing as the better strategy. Deslauriers, McCarty, Miller, Callaghan and Kestin (2019) found students in active classes learned more and reported learning less. Learners prefer what works least, and they go on preferring it after being shown the evidence. A product that optimises for what learners prefer will optimise for exactly that.

The move

The useful distinction is between two kinds of difficulty. Some effort is caused by the interface: a confusing layout, a control that does not do what it looks like it does, a diagram on one screen and its explanation on the next so the learner holds one in memory while hunting for the other. That effort is waste, and every commercial instinct about removing it is correct here too. The other kind is the effort of thinking hard about the material. Retrieving it rather than re-reading it. Attempting a problem before being shown the method. Telling two similar cases apart. That effort is the mechanism. Remove it and you have removed the learning while keeping the product.

So the job changes. The system does not need to feel effortless. It needs to make the right effort possible, visible and worth doing, and to remove only the effort that is wasted. The design skill is the same in both fields. The judgement about which kind of effort you are looking at is the whole difference between them.

A second move follows. Engagement can no longer serve as the proxy. Time on task, completion, streaks and satisfaction scores are what a commercial product measures because they are what it wants. In a learning product they still measure something, and the something is use. The measure that matters is what the learner can do a week later that they could not do before. That is slower to collect and much harder to put on a dashboard, so most learning products never collect it, and never find out.

What it produced

The first thing was a question to hold every screen against. Does this help the learner do the thinking that causes learning, or does it make the experience feel good? A screen can do both. Many do only the second, and they pass review because the second is what review was looking for.

From that one question came a short lens for reading any screen. What does the learner actually do here, as opposed to watch or read? Is the difficulty in the material or in the interface? Will the help fade as the learner improves, or become the thing they depend on? Where is the retrieval? What is the reward doing to the reason to learn? Which learner was this built for, and who is left out by that choice? And what will the learner be able to do somewhere else because of it? None of these is exotic. What is new is asking them of a product that was designed to be enjoyed.

The second thing was a change in how I hear the phrase “the learner is struggling”. I now ask what with. If the answer is the interface, fix it today. If the answer is the material, the next question is whether the struggle is productive, and that is a question about the design of the task rather than a bug in the product.

The third was a way of talking to partners from consumer software without treating their instincts as wrong. They are right, for most products. Learning is the exception, and it is an exception for reasons that fit on one figure with two crossing lines. Saying that early, with the evidence attached, has moved more decisions than any principle I have handed over.