When Fairness Meets With Reasonableness in Recommender Systems

When Fairness Meets With Reasonableness in Recommender Systems.

Contents. Recommender System Unfairness issues in recommender system Fair recommender system Sacrifice in fair recommender system Our claim & Motivation Our contribution Related work Proposed method.

RECOMMENDER SYSTEM. A recommender system utilizes algorithms to suggest relevant products, such as books, clothing, and movies, to users by analyzing their historical behaviors, preferences and interactions with products..

How Recommendation System Must Be. A person with a computer Description automatically generated Male.

How Fair Recommendation System Ensures Equal Opportunity.

Problem in Fair Recommendation System. A person with a computer Description automatically generated Male.

Scenario 1: Reasonable Treatment. Not Recommend. A cartoon of a child Description automatically generated Male.

OUR CONTRIBUTION. Novelty: First work to address unfairness as (un)reasonable treatment based on item characteristics. Introducing Reasonable Fairness Metrics: Measure treatment discrepancies between the similar treatment of users conditioned on the item’s context, rather than solely on their sensitive attributes..

RELATED WORK. 1. Demographic Parity : Same recommendation rate among protected groups 2. Equal Opportunity : Same true positive rate among protected groups 3. Predictive Equality: Same false positive recommendation rate among protected groups.

Considerable. A cartoon of a child Description automatically generated A cartoon of a child Description automatically generated A cartoon of a child Description automatically generated A cartoon of a child Description automatically generated.

THANK YOU.