We consider the problem of off-policy policy selection in reinforcement learning: using historical data generated from running one policy to compare two or more policies. We show that approaches based on importance sampling can be unfair—they can select the worse of two policies more often than not. We give two examples where the unfairness of importance sampling could be practically concerning...

We study fairness in collaborative-filtering recommender systems, which are sensitive to discrimination that exists in historical data. Biased data can lead collaborative-filtering methods to make unfair predictions for users from minority groups. We identify the insufficiency of existing fairness metrics and propose four new metrics that address different forms of unfairness. These fairness me...

In this paper, we study several TCP fairness issues in wireless local networks, both experimentally and via simulation. We illustrate the unfairness problems caused by the 802.11 MAC protocol, as well as by cross-layer interactions. Two possible solutions, namely SPM-AF queue management and LAS queue management, are proposed in the paper. These two queue management approaches are designed to al...

When does procedural unfairness result in retaliation, and why do recipients of unfair treatment sometimes pursue and other times inhibit retaliation? Five studies addressed these questions. The authors proposed and found that regulatory focus moderates retaliation against an unfairness-enacting authority: Promotion-focus participants were more likely to retaliate than prevention-focus particip...

We experimentally investigate whether individuals are more likely to engage in dishonest behavior after having experienced unfairness perpetrated by an individual with a salient group identity. Two individuals generate an endowment together, but only one can decide how to share it. They either share the same group identity or have distinct group identities. Then, they approach a task in which t...

This paper deals with coexistence issues among contention-based heterogeneous wireless networks that have different transmission power and/or coverage. We show that the existing carrier sensing multiple access mechanism, which is a prevailing contention-based protocol, results in a significant unfairness of channel sharing. We analyze the reasons of such unfairness by considering the mutual eff...

Current Differentiated Services (DiffServ) networks supporting Assured Forwarding (AF) service succeed in providing minimum rate guarantees, but fail to distribute network capacity in a fair way. To address the unfairness problem we propose simple adaptive markers operating at the network edge, whose marking function adapts to changes of the traffic mix. Extensive simulation experiments indicat...

It is well known that TCP experiences performance problems (unfairness and starvation) in multi-hop wireless networks. Such problems are often traced to the misbehavior of dynamic routing protocols in case of link failure. However, in this paper we show, through simulations, that unfairness and starvation of flows also take place with static routing; thus competition between TCP flows in a mult...

In this paper, we propose a new solution to cope with the unfairness limitations of the Distributed Coordination Function (DCF) algorithm. Indeed, in current widely deployed IEEE 802.11b Wireless Local Area Networks (WLAN), the performance of all the competing access nodes are dramatically affected once the bit rate of one station degrades. This anomaly is due to the unfairness behavior of the ...