How recommendations work quiz

Question 1

What is a recommendation system?

Accepted Answer

A computer program that suggests items a user might enjoy based on their past behaviour

Answer

A search engine that finds items by keyword

Answer

A program that only displays the most popular items on a platform

Answer

A security system that verifies user passwords before login

Question 2

What does Netflix use its recommendation system for?

Accepted Answer

To suggest movies and shows a user is likely to enjoy based on their viewing history

Answer

To set different subscription prices for different regions

Answer

To detect and block illegal video downloads automatically

Answer

To sort all available shows into alphabetical order for the browse page

Question 3

What is collaborative filtering?

Accepted Answer

Recommending items based on the behaviour of users who share similar tastes

Answer

Filtering out spam and inappropriate content from a platform's catalogue

Answer

Sorting items by colour or category to help users browse more easily

Answer

Blocking content that is too advanced for a user's age group

Question 4

What is explicit feedback in a recommendation system?

Accepted Answer

Feedback the user directly gives, such as a star rating or a 'like' button click

Answer

Feedback collected from a user's actions without them being asked

Answer

Feedback provided secretly by the app's own developers

Answer

Feedback that users share with each other in chat messages

Question 5

What is implicit feedback in a recommendation system?

Accepted Answer

Behavioural signals like watch time, clicks, or skips collected automatically without asking the user

Answer

A star rating the user directly selects from 1 to 5

Answer

A written review the user submits after finishing a movie or song

Answer

Feedback that comes from the user's social media friends

Question 6

What is a user profile in a recommendation system?

Accepted Answer

A stored collection of data about a user's past behaviour and preferences used to personalise recommendations

Answer

A profile photo and bio the user uploads when first creating their account

Answer

A list of friends the user has connected with on the platform

Answer

The unique username and password combination that secures an account

Question 7

What is content-based filtering?

Accepted Answer

Recommending items whose features closely match items the user has already liked

Answer

Recommending only the most recently added items in the platform's catalogue

Answer

Filtering out harmful or adult content to protect younger users

Answer

Recommending items that other similar users enjoyed based on shared ratings

Question 8

What is the 'cold start problem' in recommendation systems?

Accepted Answer

The difficulty of making good recommendations when a new user or new item has very little data

Answer

A software bug that causes the app to start very slowly in cold weather

Answer

The challenge of handling too many users signing up at the same time

Answer

A database error that occurs when the server restarts unexpectedly

Question 9

What is a rating matrix, and why is most of it empty?

Accepted Answer

A table where rows are users, columns are items, and cells hold ratings — most cells are empty because users haven't rated everything

Answer

A secure grid that stores encrypted user passwords for every item they accessed

Answer

A performance dashboard that displays server response times for each item in a grid

Answer

A matrix that tracks the price history of items over different time periods

Question 10

How does cosine similarity help a recommendation system find similar users?

Accepted Answer

It measures how alike two users or items are by comparing the direction of their rating vectors in mathematical space

Answer

It counts the total number of items two users have both rated to find overlap

Answer

It measures how quickly a recommendation page loads for different users

Answer

It checks whether a user meets the minimum age requirement to view certain content

Question 11

What is a 'filter bubble' and how do recommendation systems create one?

Accepted Answer

When a recommendation system only shows familiar content, trapping the user and limiting exposure to new ideas

Answer

A spam-filtering tool that removes duplicate or low-quality recommendations from a feed

Answer

A privacy setting that hides a user's watch history from third-party advertisers

Answer

A visualisation chart that shows which content categories are trending on the platform

Question 12

What is popularity bias in recommendation systems and why is it a problem?

Accepted Answer

The tendency of a system to keep recommending popular items, ignoring niche ones that specific users might prefer

Answer

When celebrities and influencers receive more screen time than ordinary content creators

Answer

When the algorithm processes requests from high-traffic users faster than others

Answer

When users give unfairly high ratings to items made by their favourite celebrities

Question 13

What is the difference between user-based and item-based collaborative filtering?

Accepted Answer

User-based recommends from what similar users liked; item-based recommends items similar to what you already enjoyed

Answer

User-based uses star ratings while item-based uses written text reviews

Answer

User-based works offline while item-based always requires an internet connection

Answer

User-based is designed for adult users while item-based is designed for children

Question 14

What is a hybrid recommendation system?

Accepted Answer

A system that combines multiple recommendation techniques, like collaborative and content-based filtering, for better overall results

Answer

A system that alternates between recommendations and advertisements on alternate screens

Answer

A single app that runs simultaneously on both Android and iOS devices

Answer

A system powered partly by human editors and partly by the algorithm equally

Question 15

Why might a recommendation system suggest products you have already purchased?

Accepted Answer

The model may not filter out already-purchased items, or similar users frequently re-buy the same product

Answer

The algorithm is designed to remind users of previous purchases to encourage re-buying

Answer

The recommendation engine was compromised and is showing random past orders

Answer

The recommendation team manually selected those items to appear on the homepage

Question 16

What does 'serendipity' mean in the context of recommendation systems?

Accepted Answer

A pleasant, unexpected recommendation that surprises the user and turns out to be something they love

Answer

An algorithmic error that causes completely irrelevant and random recommendations to appear

Answer

A feature that shows only the safest, most age-appropriate content to every user

Answer

A practice of recommending items the user has already seen many times as a helpful reminder

Question 17

What is the 'long tail' problem in recommendation systems?

Accepted Answer

The challenge of recommending the many niche items that have few ratings but matter greatly to specific users

Answer

A display error where item titles are too long to fit correctly in the app interface

Answer

The difficulty of processing large datasets that take a very long time to compute

Answer

A scrolling bug that causes the recommendation list to extend without ever ending

Question 18

What is the K-Nearest Neighbours (KNN) algorithm and how is it used in recommendations?

Accepted Answer

An algorithm that finds the K most similar users or items and uses their data to make recommendations for the target user

Answer

An algorithm that picks K completely random items from the catalogue to suggest every day

Answer

A method to block K unwanted user profiles from being able to view your account

Answer

A technique that simply counts the total number of items a user has ever rated

Question 19

What is matrix factorization and why is it useful in recommendation systems?

Accepted Answer

A technique that splits the user-item matrix into two smaller matrices of hidden factors to predict missing ratings

Answer

A technique that multiplies all user ratings by a scaling factor to normalise scores across the platform

Answer

A method to remove duplicate rows and columns from the rating database to save storage

Answer

A method to encrypt the rating matrix so that it can be stored securely on the server

Question 20

What is a feedback loop in recommendation systems and why can it be harmful?

Accepted Answer

A self-reinforcing cycle where recommended items get more clicks making the system recommend them even more, crowding out diverse content

Answer

A coding loop that continuously re-trains the model on outdated historical data

Answer

A feature that asks users for feedback after every single recommendation they click

Answer

A user interface element that replays the same recommended video until the user provides a rating

Question 21

What are 'latent factors' in a recommendation system?

Accepted Answer

Hidden numerical patterns the model learns automatically that capture user preferences and item characteristics

Answer

Features explicitly labelled by human editors in advance, such as genre, price, or release year

Answer

The total number of inactive accounts that have never rated or interacted with any item

Answer

Performance factors that measure how slowly a recommendation result appears on screen

Question 22

What ethical problem can arise when recommendation systems are used to deliver news?

Accepted Answer

Optimising for clicks or watch time can cause the system to promote sensational or misleading content, spreading misinformation and deepening polarisation

Answer

Journalists lose their jobs because the algorithm can automatically write and publish news articles

Answer

The algorithm charges users a fee every time they open a recommended news article

Answer

Too many users reading recommendations simultaneously can overload news platform servers

Question 23

What is the explore-exploit trade-off in recommendation systems?

Accepted Answer

The balance between recommending proven favourites and occasionally trying new unknown items that might suit the user better

Answer

Deciding whether to build the recommendation engine internally or purchase a ready-made solution from a vendor

Answer

Choosing between a fast but less accurate model and a slow but perfectly accurate one

Answer

Trading anonymised user data with partner platforms in exchange for access to a more diverse item catalogue

Question 24

How does deep learning improve recommendation systems compared to older techniques?

Accepted Answer

It automatically learns complex patterns from raw data like images, audio, and text to build richer representations of users and items

Answer

It makes databases physically larger so that more user data can be stored on the server

Answer

It restricts the item catalogue to only content that neural networks classify as high quality

Answer

It reduces the total number of recommendations shown on screen to prevent information overload

Question 25

What is cross-domain recommendation?

Accepted Answer

Using data from one domain, such as music, to improve recommendations in a different domain, such as movies or books

Answer

Showing identical recommendations to users in different countries and automatically translating item titles

Answer

Automatically translating item descriptions so they appear in the user's preferred language

Answer

Sharing a single recommendation feed across multiple user profiles in the same household account

Question 26

What is A/B testing and why do recommendation teams use it?

Accepted Answer

An experiment that shows two different algorithm versions to separate user groups and compares real behaviour to choose the better one

Answer

A quality check where internal testers give letter grades of A or B to each individual recommendation

Answer

A method that alternates between two recommendation algorithms for the same user every session

Answer

A line-by-line code review process that checks the recommendation algorithm for bugs

Question 27

What does it mean for a recommendation to be 'personalised'?

Accepted Answer

The recommendation is tailored to a specific user's individual history and preferences

Answer

The recommendation is updated every 60 seconds to always show the latest trending items

Answer

Every user on the platform receives exactly the same homepage recommendations

Answer

The recommendation includes the user's first name written inside the title of each suggested item

Question 28

What kinds of information does Spotify use to recommend songs to you?

Accepted Answer

Listening history, replays, skips, playlist additions, and audio features of songs like tempo and energy

Answer

Only the genre tags the user manually selected when creating their account

Answer

Primarily the user's geographical location and the current local weather conditions

Answer

The total number of followers each artist has on the platform

Question 29

What is the difference between real-time and batch recommendations?

Accepted Answer

Real-time recommendations adapt instantly to a user's current actions; batch recommendations are pre-computed and served from stored lists

Answer

Real-time uses older computers while batch recommendations always use the fastest supercomputers

Answer

Real-time recommendations work without internet; batch recommendations require a live connection

Answer

Real-time shows random items to explore new content; batch shows the most popular items on the platform

Question 30

How can a recommendation system be unfair to certain groups of users?

Accepted Answer

By learning and reproducing biases in the training data, giving poorer recommendations to underrepresented groups of users

Answer

By deliberately showing longer loading screens to users who have rated fewer items

Answer

By charging a premium price for personalised recommendations while giving only generic ones to free users

Answer

By accidentally sharing one user's private watch history with other users on the platform

Test yourself