← All simulations · Pillar 3: Sorting things into groups
Decision-Tree Builder
A decision tree is the most see-through kind of AI: it just asks a chain of yes/no questions. Here you can grow one deeper and watch it carve the space into blocks — one question is one straight cut. The region map, the flowchart of questions, and the Spectra code all stay locked together, so you can always see how the tree decides. Then drop a new fruit in and watch it fall down the questions to an answer.
Meet the parts
- A question (a split). Every box that isn’t a leaf asks one yes/no question about a single number, like “is sweetness below 0.3?”. That one question draws one straight cut across the map — everything on one side goes left, everything on the other goes right.
- Depth. How many questions the tree may ask in a row. Depth 1 makes one cut; each extra level lets the tree ask a different follow-up question inside each block, so the blocks get smaller and the boundary turns into a staircase.
- A leaf (the answer). A box with no more questions. It simply votes the most common kind of fruit that landed in it — that vote is the tree’s final answer.
How to read it
- The region map (top): sweetness goes left→right, size goes bottom→top. The shading is the tree’s answer for every spot; each straight line is one question. Dots are the practice fruit; the big ringed dot is the fruit you’re testing.
- The flowchart (lower): the very same tree drawn as questions. Follow yes to the left child and no to the right, down to a leaf.
- Watch it grow ▶ adds one level at a time — a new cut appears and more practice fruit ends up on the right side. The little curve shows how the score climbs with depth.
- Drop it in ▼ sends your test fruit down the questions, lighting each box on its path until it reaches a leaf.
Try this
- Start at depth 1: one cut can’t wrap around both fruits, so it’s stuck well below 100%. Slide to depth 2, then 3, and watch the staircase form.
- Push depth all the way up. Notice the score reaches 100% — but the blocks get tiny and start hugging single dots. That’s the tree starting to memorise instead of learn (overfitting).
- In the code, change
depth: 3to another number — the picture above follows. - Move the sweetness and size sliders to a spot near a boundary, hit Drop it in, and see which leaf claims it.
- Hit Open in the Studio to keep going with the very same tree as Spectra code.
Safe by design
Everything here runs in your browser on small, made-up data. There is no upload, no internet from the model, and nothing about you is collected — the same safety-by-absence rule as the rest of the platform.