my BBC
Personalisation Recommendations & Desk Research.
myBBC wanted to get a better understanding of the reasons users come to recommenders systems and their behaviour. For this project, I collaborated with a project lead and a UX Designer as my stakeholder whom I presented my work.
Understand.
I focused on users' motivations rather than similarity algorithm capabilities. Therefore, I conducted a Desk Research in three different areas; academia, industry and previous work within the BBC. BBC work focused mainly on similarity, trust and accuracy in recommender systems. However, most important academic papers and technical articles went beyond 'accuracy' and 'trust' as a purpose. Research suggests that user satisfaction is not always related to higher recommendation accuracy but in user experience as a whole.
Create.
Desk Research allowed me to understand the problem and develop five design key requirements for a personalised recommender engine.
Design for a variety of user needs: Users are looking for serendipity, novelty and diversity, apart from the similarity in results.
Design input around users goals: Human motivation theories suggest that users input information when the output at present or in future will help them achieve their aims.
Design recommender system prioritising on Gestalt principles, rather than on list item accuracy: Users are driven by their task at hand, experiencing the resulting list as a whole.
Design the recommender system as a character that is shaped by that "give and take" with the user. Users prefer to interact with the recommenders in a conversational style. To them, recommenders have personalities.
Evaluate the recommender system measuring users overall satisfaction, using Pearl Pu's framework (2011).
Deliver.-
The research resulted in a presentation of findings on novelty, conversational interaction UI, motivation and evaluation framework. It helped my BBC to find out more about the users' motivations and information needs from a recommender system across different projects. The BBC R&D team warmly received the findings as more future-oriented.