# How Music Apps Learn What You Don't Want in Playlists > This patent describes how a music streaming service learns what kinds of songs or artists a user dislikes for their playlists by tracking what they repeatedly ignore, then uses that information to avoid recommending similar things in the future. - **Patent:** US 12277178 - **Original title:** Media content item recommendation system - **Owner:** Spotify AB - **Granted:** 2025 - **Status:** Active - **Times cited:** 0 - **Field:** software, telecommunications, consumer_electronics, ai_ml ## What it does The system first shows a user a playlist and offers recommended songs to add to it (Claim 1). It then keeps track of how often songs with specific traits, called 'attributes' (like a certain artist, per Claim 3 and 4), are suggested for a playlist but the user doesn't pick them (Claim 1). If a user consistently ignores songs with a particular attribute, the system gives that attribute a lower score (Claim 1). This score is then used to reduce the chance of recommending other songs with that same attribute for playlist additions in the future (Claim 5). For example, if you keep skipping recommendations for songs by 'Artist X' when building a playlist, the system learns to show you fewer songs by 'Artist X' or similar artists for that playlist. ## What it does NOT cover - Does not cover recommendations based solely on what a user likes or plays frequently, without considering ignored suggestions. - Does not cover general content recommendations outside the specific context of adding to an existing playlist. - Does not cover systems that track ignored individual songs without linking that behavior to broader 'attributes' like artist or genre. - Does not cover adjusting recommendations based on explicit 'dislike' buttons, only implicit ignoring by not selecting. - Does not cover recommendations where the user is not presented with a 'set of recommended media content items for addition to the playlist'. ## The clever bit The novelty lies in using repeated *non-selection* of recommended items for *playlist inclusion*, tied to specific *attributes*, to actively reduce the likelihood of future recommendations with those attributes. Most recommendation systems heavily weigh positive signals; this patent focuses on leveraging implicit negative feedback. ## Real-world examples 1. Spotify's 'Enhance' feature for playlists 2. Playlist creation tools in music streaming applications 3. Personalized radio stations that adapt based on skipped tracks ## Why it matters Personalization is critical for media streaming services. This patent describes a method for refining recommendations by learning from negative signals—specifically, what users repeatedly choose *not* to add to their playlists. This approach is crucial for improving user satisfaction and engagement by reducing irrelevant suggestions. Better recommendations can lead to users spending more time on a platform and discovering new content they genuinely enjoy, while avoiding content they dislike. ## Frequently asked questions ### What does How Music Apps Learn What You Don't Want in Playlists cover? This patent describes how a music streaming service learns what kinds of songs or artists a user dislikes for their playlists by tracking what they repeatedly ignore, then uses that information to avoid recommending similar things in the future. ### Who owns patent US 12277178? Spotify AB owns this patent, granted in 2025. ### When does this patent expire? This patent is expected to expire on May 23, 2043, when the invention enters the public domain. ### What problem does this patent solve? Personalization is critical for media streaming services. This patent describes a method for refining recommendations by learning from negative signals—specifically, what users repeatedly choose *not* to add to their playlists. This approach is crucial for improving user satisfaction and engagement by reducing irrelevant suggestions. Better recommendations can lead to users spending more time on a platform and discovering new content they genuinely enjoy, while avoiding content they dislike. ### What does this patent NOT cover? Does not cover recommendations based solely on what a user likes or plays frequently, without considering ignored suggestions. **Full plain-English explainer:** https://patentbrief.org/patent/us/12277178/media-content-item-recommendation-system **Original patent:** https://patents.google.com/patent/US12277178 --- _Source: PatentBrief — https://patentbrief.org. Patent facts are from public records; the plain-English explanation is PatentBrief's._ ## Related patents Semantically similar inventions in the PatentBrief corpus: - [How to Play Any Media Playlist by Converting it to a Standard Format](https://patentbrief.org/patent/us/6990497/dynamic-streaming-media-management) — This patent describes a system that takes media playlists in various formats, converts them into a single standard format, and then streams the referenced content, even allowing for dynamic changes during playback. - [How Facebook Uses Deep Learning to Predict What You Might Like](https://patentbrief.org/patent/us/10402750/automl-neural-architecture-search) — A method for training AI models to recommend new content by comparing a user's past interactions with unseen items in a social network. - [How Eventbrite Recommends Events Based on Your Social Network](https://patentbrief.org/patent/us/8700540/facebook-messenger) — A system that suggests events to you by analyzing your social media connections and your past attendance history to see what your friends are doing. - [How Apps Compare Local Prices and Ratings for Specific Items](https://patentbrief.org/patent/us/10127595/airbnb-dynamic-pricing-smart-pricing) — A system that helps users find and order a specific product from nearby stores by ranking them based on price, ratings, and item attributes like calorie count. - [Smart Ranking of Emails and Files Based on How You Click](https://patentbrief.org/patent/us/6370526/google-adwords-pay-per-click) — IBM's 1999 patent on automatically sorting lists of items, like emails, by watching which ones you click first and updating a mathematical model of your preferences in the background.