Executive Summary
Recommendations for online information is one of the most common ways to deal with large amounts of data. To keep users interested and engaged, YouTube recommends videos it thinks people would like to watch based on a variety of information it collects about user behavior. In this study, we analyze how people perceive these recommendations and how successful they find them to be. We further investigate how people go about making their decisions when it comes to selecting a recommendation and if people even prefer a personalized list over a non personalized list of videos. We conducted a user experience evaluation to gather qualitative data and then analyzed the responses through a thematic analysis to find themes in how people react to the YouTube recommender system. Our results suggest:
Novelty is not always desirable
People often struggle with a flood of irrelevant suggestions
Furthermore, we discovered that people tend to judge videos based on the title and thumbnail that is presented in addition to taking the length of the video into account. Overall, our participants were able to distinguish between personalized recommendations and non-personalized recommendations.
Want to read rest of the Article?
Role in Project
UX Research
Researched about recommendation systems, applied thematic analysis and made some qualitative conclusions.
User Interviews
Designed the interview protocol and conducted a qualitative user study using the "think-aloud" protocol to understand how user's use Youtube's Recommendation System.
Process
-
Researched Statistics
Researched youtube statistics and academic papers about recommendation systems.
-
What are our Research Questions and Target Audience?
How successful do students find the personalization of Youtube’s mobile app recommendations? How do people make their decisions about which recommendation to select in the context of 15 minutes to spare? Do people even prefer a personalized form of the recommendations over a non personalized form and why?
-
Framed user-interview Protocol
As a team, we discussed about possible questions to ask during user-interviews and hence decided to apply the Think-aloud protocol.
-
Conducted User-interviews
Initiated by obtaining oral consent, informed that I will be taking notes and reminded them to 'think-aloud' and collected information for thematic analysis.
