Fairness of Exposure in Stochastic Bandits

               Lequn Wang 1 Yiwei Bai 1 Wen Sun 1 Thorsten Joachims 1

            Abstract                 While maximizing user responses may arguably be in the
                                 interest of the platform and its users at least in the short term,
   Contextual bandit algorithms have become widely       there is now a growing understanding that it can also be
   used for recommendation in online systems (e.g.       problematic in multipl ...