Mitigating Filter Bubbles within Deep Recommender Systems
This addresses filter bubbles that isolate users in social media and e-commerce, but it is incremental as it builds on existing methods.
The paper tackled the problem of filter bubbles in deep recommender systems by characterizing and mitigating the effect without compromising accuracy, using TracIn to analyze influences and retraining the system.
Recommender systems, which offer personalized suggestions to users, power many of today's social media, e-commerce and entertainment. However, these systems have been known to intellectually isolate users from a variety of perspectives, or cause filter bubbles. In our work, we characterize and mitigate this filter bubble effect. We do so by classifying various datapoints based on their user-item interaction history and calculating the influences of the classified categories on each other using the well known TracIn method. Finally, we mitigate this filter bubble effect without compromising accuracy by carefully retraining our recommender system.