«Personalized Recommendations will replace the navigation grid on Netflix»

So far, recommender systems have been mostly evaluated on accuracy → one-click analysis We extend this evaluation towards sequences of dependent clicks →multiple-click analysis RESULTS Recommendation networks are poorly navigable, but explorative and variable scenarios are better supported. Collaborative Filtering produces more navigable networks than content-based recommendations. More attention to recommendation net-works is needed to improve navigability, e.g., by specifically replacing some rec-ommendations. RESEARCH QUESTIONS 1) How well are recommendation net-works suited for navigation and ex-ploratory search? 2) What is the influence of parameters (e.g., recommendation algorithms and the number of recommendations shown) on navigability? REFERENCES [1] M. J. Bates. The design of browsing and berrypicking techniques for the online search interface. Online Infor-mation Review, 13(5):407–424, 1989. [2] J. Kleinberg. Complex networks and decentralized searc