AI Personalization Engines Quiz

Answers and explanations appear as you take the quiz.

1. What does collaborative filtering primarily rely on?
   • Patterns across many users, e.g. 'people who liked X also liked Y'
   • Only the text description of each item
   • The user's device battery level
   • A fixed editorial top-10 list
2. What is the 'cold-start problem' in personalization?
   • Servers taking too long to boot
   • Having little or no data about a new user or new item
   • Recommendations only working in winter
   • Users logging out frequently
3. In a typical two-stage architecture, what does the first (candidate-generation) stage do?
   • Precisely ranks only ten items
   • Writes the final product descriptions
   • Quickly narrows millions of items down to a few hundred
   • Deletes inactive user accounts
4. Why might a recommender add a diversity objective to a ranked list?
   • To guarantee every item receives equal clicks
   • To remove all user-history signals
   • To make every recommendation random
   • To avoid filling the list with near-duplicate items and broaden the alternatives shown
5. What are 'embeddings' in a recommendation system?
   • Vectors representing users and items so closeness signals relevance
   • Physical chips in the server
   • Cookies stored in the browser
   • Email subject lines
6. Why is implicit feedback (clicks, watch time) often weighted heavily over explicit star ratings?
   • It is easier to fake
   • It is far more abundant and reflects actual behavior
   • It is required by law
   • Star ratings are illegal
7. What does content-based filtering use to recommend items?
   • Only other users’ passwords
   • Random item order
   • Features of items a user has liked or interacted with
   • The device’s battery history
8. What does normalized discounted cumulative gain (NDCG) evaluate?
   • The number of users who created accounts
   • The model’s training duration
   • Whether an item description is grammatically correct
   • The quality of a ranked list, giving higher weight to relevant items near the top