Popularity and Demographic Biases in Recommender Systems

Explore the impact of popularity and demographic biases in recommender systems through this conference talk from FAT* 2018. Delve into the core question of ensuring recommendation algorithms work effectively for all users. Examine various evaluation strategies, datasets, and algorithms used to assess and address biases. Learn about initial findings related to profile size, resampling techniques, and methods to mitigate popularity bias. Understand the limitations of current approaches and discover upcoming workshops in this field. Gain insights into systematic differences in recommender systems and their implications for fairness and effectiveness across diverse user groups.

Introduction
Core Question
What is a Recommender
How do we make sure systems are good for everyone
What are we looking at
Vocabulary
Valuation Strategies
Data
Datasets
Results
Algorithms
Initial Results
Profile Size
Resampling
Popularity Bias
Fixing Popularity Bias
Limitations
Upcoming Workshops
Questions
Recommendations
Systematic differences