Author: Andreas Fink
Supervisor: Julia Neidhardt
Abstract
Early stages of user interactions pose a significant challenge for recommender systems, particularly due to the cold start problem. This issue arises when new users or items lack historical data, making it difficult to generate personalized recommendations. This thesis investigates the effectiveness of four visual preference elicitation methods – Swipe, Rating, Two-Items, and Four-Items – for constructing an initial preference profile during the onboarding process of a mobile leisure activities recommender system. For this purpose, a browser-based survey prototype was developed, and these methods were evaluated in terms of completion rate, time efficiency, usability, and profile accuracy. The study included 382 participants who completed the survey. The results show clear trade-offs between ease of use and the quality of the resulting preference profile. The Swipe method achieved the highest scores in usability and completion rate but resulted in the least accurate preference profiles. The Rating method produced the most accurate profiles but required more time and showed a lower completion rate than the Swipe method. The Two-Items method showed a balanced performance in terms of time efficiency and profile accuracy but had the lowest completion rate. While the Four-Items method performed well in terms of completion, its profile accuracy was lower than that of the Rating method and the Two-Items method. Additionally, the completion time was significantly longer compared to the other methods. This thesis contributes to the mitigation of the cold start problem in recommender system and provides valuable insights into the design of user-friendly and effective preference elicitation methods in the context of a mobile leisure activities application.