Which approach?

A structure in which students sort mixed problems by the approach that solves them, without working them out. Four steps: everyone silently writes next to each problem which approach it needs; in pairs, first A and then B says what gave it away ("I recognise it by …"); together the pair writes one recognition rule per type ("If there is … in it, then …"); and a drawn pair reads its rule out, which you type live into that type's tile on the board while the class adds to it. The structure trains recognising the type of problem, the part that most often goes wrong on mixed tests.

With the default settings the whole structure takes ± 16 minutes, briefing and look-back included.

Step by step with the defaults: three minutes to sort, two to compare (after a minute the turn passes from A to B by itself), two for the rules and four for the board. Whoever prepares the structure sets each step; a step set to 0 is skipped. The board is done when every rule is up, not when the time runs out, so set it to move on your click. The solving step is off by default; switch it on and one step is added.

Because calculating crowds out choosing. A student who calculates works two problems in three minutes and chooses twice; a student who only chooses does it eight times. And choosing is what an ordinary practice set never trains, because the unit has already made that decision for the student. If you want to show that a good rule also leads to the right answer, switch on the fifth step: one problem of each type, solved after the board.

Six to eight problems of two or three types is a good start; the setup allows four to twelve problems and two to four types. Make sure each type appears at least twice, otherwise there is nothing to compare within a type. More than three types makes the rules on the board long and the conversation in the pair shallow. For students still learning English, the Language learners variation sets up five problems and two types.

From your own textbook, or have them written. With a free account, the built-in AI writes them in the setup: give your topic and grade in the bar above the Problems field, and you get two or three types mixed together, which you read through before you start. If they are already in your book, take a photo of the page with the camera next to that bar, and the problems land in the field. Or leave the field empty: then they stay in the book and you only choose how many. Prefer Microsoft Copilot? The same screen has the prompt ready: "Write 8 short problems about [topic] for students aged [age], of two or three types that each need a different approach, mixed together. Do not name the approach in the problem. One problem per line, no numbering." Paste the answer into the Problems field.

One that names something you can see before you start calculating, and that also works on a problem you have not seen. "If there is a discount or sales tax in it, then percent" is a start; "if it asks for a part of a whole and the whole is a hundred percent, then percent" is better. "Because it is a percent problem" is not a rule: that is the name again. On the board, test each rule with one question to the class: does it also fit the other problems of that type? If a feature you find important is missing, add it right there.

Yes, wherever students know a rule but do not see when it applies. In English language arts, comma rules are the example: most comma errors are recognition errors, not punctuation errors. With English learners it is the tenses, in history the math with centuries and dates, in science the formula that fits a situation. The structure stays the same; only the problems and the names of the types change.

Because otherwise the same pairs read every time, and the rest never needs a rule. With the draw, nobody knows who will speak for the class, so every pair has a rule ready. The screen says so too: "Nobody knows yet which pair speaks for the class. That is why every pair has a rule ready." You draw again for each type, so with three types, three pairs get the floor.

Practicing with problems of different types mixed together, instead of one block per type. In the research by Doug Rohrer and colleagues, students who practice this way score clearly higher on a later test, even though the practice itself feels harder and produces more errors. The explanation is the choice: in a block you already know the approach, in a mix you have to find it for every problem. Which approach? takes just that choice and makes it audible, so you hear what the class recognises a type by.

In what gets practiced. In Pair coach, one student works a problem out loud while the partner watches: the approach is given, carrying it out is the practice. In Which approach? nothing is carried out: choosing the approach is the practice. They work well back to back: first sort with Which approach?, then work one problem of each type with Pair coach.

Which approach? is a LudensFlow structure built on two lines of research. The first is the work on interleaved practice by Doug Rohrer and colleagues at the University of South Florida (2007, 2015, 2020): mixing practice problems improves learning, mainly because students then have to choose. The second is the classic sorting study by Michelene Chi, Paul Feltovich and Robert Glaser (1981): novices group problems by what they can see, experts by the principle underneath. The build (alone first, then in pairs, then with the class) and the drawn pairs come from cooperative learning: everyone makes their own contribution, and everyone can be called on.