Master thesis student
Thesis topic: The effects of personal characteristics and level of detail on the perception of explanations in a recommender system
Related project: RIMA
Supervisor: M. Sc. Mouadh Guesmi
Thesis duration: 11 / 2020 - 05 / 2021
Recommender systems have been increasingly used in online services such as Amazon, Netflix, or YouTube. However, recommender systems are often presented as a 'black-box', where insights into the logic or justification of the recommendations are hidden to the users. As a result, users increasingly demand for more transparent systems. Therefore, the concept of explainable recommendations has attracted much attention from researchers over the past years. There is one important aspect that designers of explainable recommender systems need to reflect upon: How much information should the system provide? It must be considered that, because of interpersonal differences, not every user requires the same (amount of) information. There are many factors which could influence the perception of a recommender system and its explanations, for instance the user's personality, attitudes, beliefs, and cognitive abilities. In addition, demographic variables like age, gender, education, and profession could play an important role as well. Hence, one approach to ensure the acceptance of different groups of users is to provide personalized explanations, which allows users to select the most appropriate explanation based on their preferences (e.g. whether or not to see the explanations, see different levels of explanation detail).
The aim of this master thesis is to evaluate a recommender system with different levels of explanation detail through a user study and qualitative interviews. The users will rate the explanations based on different criteria, such as the perceived transparency or trust. Moreover, we want to find out whether there are differences in the perception of explanations when taking personal characteristics into account.