Toward Fair Recommendation in Two-sided Platforms

Many online platforms today (such as Amazon, Netflix, Spotify, LinkedIn, and AirBnB) can be thought of two-sided markets with producers customers goods services. Traditionally, recommendation services in these have focused on maximizing customer satisfaction by tailoring the results according to personalized preferences individual customers. However, our investigation reinforces fact that such customer-centric design may lead unfair distribution exposure producers, which adversely impact their well-being. On other hand, a pure producer-centric might become As more people are depending earn living, it is important ensure fairness both In this work, mapping fair problem constrained version fairly allocating indivisible goods, we propose provide guarantees for sides. Formally, proposed {\em FairRec} algorithm Maxi-Min Share ($\alpha$-MMS) Envy-Free up One Item (EF1) Extensive evaluations over multiple real-world datasets show effectiveness ensuring while incurring marginal loss overall quality. Finally, present modification FairRec (named FairRecPlus) at cost additional computation time, improves performance customers, maintaining same guarantees.