Design of recommender systems
Download as PPT, PDF101 likes18,144 views
The document discusses the evolution of recommender systems from 2001 to 2006 and design strategies for improving user experience. It notes that early systems focused on predicting items like movies or music a user may like, while newer social systems in 2006 helped users find information by facilitating social connections using tags and user-generated content. Key recommendations include making systems personally useful before providing recommendations, making participation and the process social, providing instant gratification, cultivating independent crowds, and balancing public and private sharing options.
![Questions? [email_address] URLs www.uzanto.com www.slideshare.net](https://image.slidesharecdn.com/design-of-recommender-systems-15305/85/Design-of-recommender-systems-47-320.jpg)
Design of recommender systems
- 1. Design Strategies for Recommender Systems Rashmi Sinha www.uzanto.com Jan 2006, UIE Web App Summit
- 2. What are Recommender Systems? Circa 2001 Systems that attempt to predict items, e.g., movies, music, books, that a user may be interested in (given some information about the user's profile) e.g., Amazon – people who liked this book also liked, Netflix recommendations Circa 2006 Systems that help people find information that will interest them, by facilitating social / conceptual connections or other means… Pandora, Last.fm
- 3. Designing different finding experiences Some experiences guide user, others just point in a general direction Desired experience depends on user task, time constraints, mood etc. There’s more than one way to get from here to there…
- 4. User experience in search/browse interfaces More controlled experience Every movement (forward, making a turn) is a conscious choice System should provide information at every step If user takes wrong turn, go back a step or two / start again Like driving a car…
- 5. User Experience with Recommender Systems User has less control over specifics of interaction System does not provide information about specifics of action More of a “black box” model (some input from user, output from systems) Like riding a roller coaster…
- 6. Recommender Systems Circa 2001
- 7. what movies you should watch… (Reel, RatingZone, Amazon) what music you should listen to… (CDNow, Mubu, Gigabeat) what websites you should visit… (Alexa) what jokes you will like… (Jester) where to go on vacation (TripleHop) & who you should date… (Yenta) I know what you will read next summer!
- 8. A technological proxy for a social process “ I think you would enjoy reading these books…” Friends / Family Ref: Flickr photostream: jefield Ref: Flickr-BlueAlgae Ref: Flickr-Lady_Strathconn What should I read next?
- 9. Interaction paradigm “ Books you might enjoy are…” Output: Input: Rate some books Ref Flickr photostreams: anjill154 & rossination What should I read next?
- 10. Meg & James: correlation = .52 How collaborative filtering algorithms work Recommendations For Meg Lets find a book for Meg!
- 11. Input : Motivating users to give input (to feed collaborative filtering algorithms) System : Making good, useful recommendations (effectiveness of algorithm) Output (Recommendations) : Presenting recommendations quickly enough but not too quickly (knowing when to say “I can’t recommend”) Generating trust that system understands user tastes Providing enough information about each item Challenges of Recommender System design
- 12. Domain differences drive design Form of sample (song clip vs. product description vs. full text article) Genres: how fixed and predictable are they? Frequency of updates (e.g., news & other fast-flowing content) Commerce vs. taste exploration vs. info-seeking
- 13. Some observations & design principles
- 14. Trust is crucial Users think recommender systems have personalities First impressions are crucial Does system understand me? Should I act on its recommendations? Two different approaches: Amazon offers affirming experience: familiar items may be correct but not as useful (not new information) MediaUnbound : less familiar, so more salient and possibly serendipitous, but less likely be acted upon Source: Sean McNee, John Riedl, Joseph Konstan, CHI Proceedings 2006 “ Making Recommendations Better: An Analytic Model for Human-Recommender Interaction”
- 15. Make system logic transparent Users want to understand why an item was recommended to them To decide whether to accept recommendation Explaining recommendations Identify the input for particular recommendation
- 16. How to motivate participation Design principle: Easy & engaging process for giving input (MediaUnbound) Ask at the right moment (Netflix)
- 17. Give users control… Design Principle: Offer filter-like controls for genres/ topics. Ask how familiar recs should be
- 18. Provide detailed info about recommended items Design principle: Provide clear paths to detailed item information and community feedback such as Reviews Ratings by other users Sample of item
- 19. The unfulfilled promise of Recommender Systems Some very popular systems (Amazon & Netflix) Overall, recommender systems lost steam—nowhere near as popular as search. Data sparseness (unlike search which builds on preexisting data – hyperlinks) Cold start problem Interface issues Gaming the system / spam etc. Hard to understand and control Lacked a larger purpose; an end in themselves Source: Paolo Massal and Bobby Bhattacharjee, Proc. of 2nd Int. Conference on Trust Management, 2004 “ Using Trust in Recommender Systems: an Experimental Analysis”
- 21. What’s happened in the interim? Social networking systems (Friendster, Orkut, LinkedIn, MySpace) Blogs, Wikis Tagging / folksonomies Google AdSense YouTube Rich interfaces (AJAX / Flash) People read, write, play, share pics, videos on the web. They live their lives on the web.
- 22. Pandora as a textbook example of recommender design principles
- 23. Characteristics of Pandora Rich interface makes experience seamless Starts giving results with one click Puts user in control of recommendation Takes a conversational tone Transparent logic Generates trust Problems Not scalable approach Not social approach: feels like a machine doing thinking for me
- 24. Last.fm: a social approach to recommendations
- 25. Exploring music at Last.fm
- 26. Characteristics of Last.fm Quick start, friendly interface Multiple points of entry: charts, tags, users, new items - not just what system recommends for you Focus on social approach Listen to other users’ radio stations (Friends, Neighbors, Groups) Read journals Chat on message boards Highlights contributions to system: your radio station is available to others
- 28. What do these systems have in common? User-generated content: mass participation & social sharing User- curated content: tags, collections etc. Harnessing wisdom of crowds Granular addressability of content The long tail: making the esoteric more findable Incorporating social networks Rich user experience Not all work: elements of fun and play Tim O’Reilly, “What is Web 2.0: Design Patterns and Business Models for the Next Generation of Software”
- 29. A revolution in RS user experience User interacts with algorithm to get recommendations System may use aggregated data about other users (via collaborative filtering algorithms). That data is not directly accessible to all Centered on completing a finding task or making sales User interacts with other users, their content and tags to find information & connect with people Frequently tag-based Data from other users is exposed and updated in real-time Succeeds by building a social web, making it more like an ongoing conversation than a transaction 2001 2006 Intelligent Agents Information & Social Hubs
- 30. User experiences for finding
- 31. User experience with social recommender systems Move at a slower pace Get the lay of the land, experience surroundings Choose paths – what is promising, what sights lie on the way, how well worn. Easy to change directions, change paths, create your own path Flickr photostream: soundfromwayout
- 32. Design Principle 1: Make system personally useful (before recommendations) System should serve other useful purpose before it starts personalizing Portable storage (photos, bookmarks) Aggregate popular news stories & feeds Offer vehicle for trendsetters / trendspotters Provide a discussion forum Personalize once system has user data Solves input problem of early RS
- 33. Del.icio.us is useful from saving first link
- 34. Design Principle 2: Make system participatory Bite-sized self-expression Artistic expression (Flickr, YouTube) Humor (YouTube) Beyond rating items – contributions of tags, comments, items Articles Photos
- 35. Different types of participation Social software sites don’t require 100% active participation to generate great value. Implicit creation (creating by consuming) Remixing—adding value to others’ content Source: Bradley Horowitz’s weblog, Elatable, Feb. 17, 2006, “Creators, Synthesizers, and Consumers”
- 36. Design Principle 3: Make participatory process social Real-time updating makes it feel more like a conversation; sense that others are out there User profiles and photos put a human face on the system interactions Spotback
- 37. What people are doing on Digg
- 38. Design Principle 4: Instant gratification Provide personalized recommendations as soon as a user provides some input Pandora: one song instant radio station Spotback: one article rating instant articles of interest Note: need lots of user data for this to work well (cold start problem emerges again?)
- 39. Design Principle 5: Cultivate user independence Prevent mobs, optimize the “wisdom of crowds”
- 40. Cultivating wise crowds Four conditions Cognitive Diversity Independence Decentralization Easy Aggregation
- 41. Design Principle 6: Provide access to long tail, keep content fast moving Make “long tail” accessible Recommend lots of different stuff (not just most popular) Top 100 lists Keeps recs from getting stale Use time as a dimension in system design Enable fast movement. Rise to top. Get displaced. e.g., “what’s fresh today” e.g., Slideshare popularity model
- 42. Design Principle 7: Expose metadata, make it linkable Exposing tags and user lists Enable “pivot browsing” Every piece of content should have a unique, easily guessed URL.
- 43. Design Principle 8: Provide balance between public & private People can be willing to share a lot if they get the right returns Allow users to: Filter by topic/category Indicate “more like this” and “no more like this” Delete items from reading history or reset profile completely Privacy settings on Flickr
- 44. Problems of early Recommender Systems addressed Motivating participation Giving users fine-grained control Making item information available Making recommendations transparent
- 45. So what’s left to solve? Possible problems: Mob rule (ends up recommending “lowest common denominator items”) Trust issues: why should I trust another user, or the community as a whole? Degree of serendipity to allow; methods for adjusting this setting
- 46. Things to try at home! Create an account on myspace.com Read Emergence, Wisdom of Crowds Play a Multiplayer Online Game (WOW, Second Life) Play with an API (try GoogleMaps API) Try a mobile social application (DodgeBall) Ask your friends what they find “fun” on the web
- 47. Questions? [email_address] URLs www.uzanto.com www.slideshare.net