Build · Grades 8-12
Recommendation Engine
Build a cosine-similarity recommender and see how Netflix and Spotify choose what to show you next.
The AI idea behind it
A recommendation engine predicts what you will like next from what you liked before. One common approach compares items by their features: if the things you rated highly share tags such as sci-fi and exciting, items with similar tags score higher.
The similarity is a number, often calculated with cosine similarity, which measures how closely two lists of features point in the same direction. Streaming services, shops and social feeds all run on recommenders.
How it works in the app
- Rate a set of titles.
- The engine finds the tags shared by the titles you rated highly, then scores new titles by how closely they match, shown as a % match with the reason ("Because you liked action + sci-fi").
- A short explanation walks through the four steps it took. Rate again with different scores and the list changes.
Try this
Whose feed is it?
- Rate the titles honestly and write down your top three recommendations.
- Rate them again as someone with very different taste, such as a younger sibling or a grandparent, and compare the lists.
- Find a title you would enjoy that never appears in your list, and work out why.
What it shows: Recommenders show you more of what you already like. That is useful, but it can narrow what you see, a pattern often called a filter bubble.
Where it fits in the curriculum
More Build tools
Try Recommendation Engine in the app
One of the creation tools in LittleAIMaster for Grades 6-12.