Fairness Index Research Articles

Letter from Trinidad and Tobago: No final battles in the weary war on COVID‐19?

Letter from Trinidad and Tobago: No final battles in the weary war on <scp>COVID</scp>‐19?

Legal perspective on possible fairness measures – A legal discussion using the example of hiring decisions

With the increasing use of AI in algorithmic decision making (e.g. based on neural networks), the question arises how bias can be excluded or mitigated. There are some promising approaches, but many of them are based on a ”fair” ground truth, others are based on a subjective goal to be reached, which leads to the usual problem of how to define and compute ”fairness”. The different functioning of algorithmic decision making in contrast to human decision making leads to a shift from a process-oriented to a result-oriented discrimination assessment. We argue that with such a shift society needs to determine which kind of fairness is the right one to choose for which certain scenario. To understand the implications of such a determination we explain the different kinds of fairness concepts that might be applicable for the specific application of hiring decisions, analyze their pros and cons with regard to the respective fairness interpretation and evaluate them from a legal perspective (based on EU law).

Efficient Load Balancing Algorithm in Long Term Evolution (LTE) Heterogeneous Network Based on Dynamic Cell Range Expansion Bias

The traditional scheme for load balancing in a homogeneous Long Term Evolution (LTE) Network where User Equipment (UEs) associate to a node with the strongest received signal strength is not practical for LTE Heterogeneous Network (LTE HetNet) due to power disparity between the nodes. Therefore, dynamic Cell Range Expansion (CRE) based load-balancing schemes were employed by several scholars to address the challenges in the LTE HetNet. However, the fairness index in achieving the desired average user throughput and UE offloading effect is relatively low. In this work, an efficient load-balancing algorithm for LTE HetNet based on dynamic Cell Range Expansion (CRE) was developed to improve the fairness of the network for the desired throughput and UE offloading effect. The simulation results achieve a throughput gain improvement of up to 11%, while the fairness index improves by 6% compared to the existing algorithm. Further, the UEs offloading effect shows a significant improvement of 3% relative to the existing algorithm.&#x0D; Keywords: Fairness Index; Cell Range Expansion; Load Balancing; LTE Heterogeneous Network; Throughput

Cluster analysis of a large dataset of patients with Lennox-Gastaut syndrome

ObjectivesWe applied the Two-Step cluster analysis on a large cohort of patients with Lennox-Gastaut syndrome (LGS). We hypothesized that there are distinct clusters of patients with LGS based on their clinical characteristics. We also studied whether the seizure outcome differs between these patient clusters. MethodsThis was a longitudinal study of a prospectively developed database. All patients with a diagnosis of LGS were studied at the outpatient epilepsy clinic at Shiraz University of Medical Sciences, Shiraz, Iran, from 2008 until 2020. The Two-Step cluster analysis (Schwarz's Bayesian Criterion: BIC) was applied to the whole dataset considering the demographic data, seizure types, and EEG data. In the next step, the seizure outcome was compared between the clusters of patients. ResultsTwo hundred and fifty-eight patients were studied. The Two-Step cluster analysis showed that there were three clusters of homogeneous subgroups of patients with LGS, with a fair silhouette measure of cohesion and separation. The seizure outcomes were significantly different between the clusters. In cluster 1, three out of 35 patients (8.6%) were seizure-free; in cluster 2, 13 out of 34 patients (38.2%) were seizure-free; and in cluster 3, nine out of 49 patients (18.4%) were seizure-free (p = 0.009; df: 2). ConclusionThe Two-Step cluster analysis could identify three distinct clusters of patients with LGS; these data are helpful in providing appropriate prognosis and counselling for patients and their care-givers. Future studies should also investigate the effects of etiology in cluster analysis of patients with LGS.

Collaborative Computation Offloading and Resource Allocation in Multi-UAV-Assisted IoT Networks: A Deep Reinforcement Learning Approach

In the fifth-generation (5G) wireless networks, Edge-Internet-of-Things (EIoT) devices are envisioned to generate huge amounts of data. Due to the limitation of computation capacity and battery life of devices, all tasks cannot be processed by these devices. However, mobile-edge computing (MEC) is a very promising solution enabling offloading of tasks to nearby MEC servers to improve quality of service. Also, during emergency situations in areas where network failure exists, unmanned aerial vehicles (UAVs) can be deployed to restore the network by acting as Aerial Base Stations and computational nodes for the edge network. In this article, we consider a central network controller who trains observations and broadcasts the trained data to a multi-UAV cluster network. Each UAV cluster head acts as an agent and autonomously allocates resources to EIoT devices in a decentralized fashion. We propose model-free deep reinforcement learning (DRL)-based collaborative computation offloading and resource allocation (CCORA-DRL) scheme in an aerial to ground (A2G) network for emergency situations, which can control the continuous action space. Each agent learns efficient computation offloading policies independently in the network and checks the statuses of the UAVs through Jain’s Fairness index. The objective is minimizing task execution delay and energy consumption and acquiring an efficient solution by adaptive learning from the dynamic A2G network. Simulation results reveal that our scheme through deep deterministic policy gradient, effectively learns the optimal policy, outperforming A3C, deep <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$Q$ </tex-math></inline-formula> -network and greedy-based offloading for local computation in stochastic dynamic environments.

Bin-packing based offline dynamic bandwidth and wavelength allocation algorithms for power efficiency in Super-PON

Enduring the rapidly growing demand for high data rates is the main challenge for the current network providers. Super passive optical network (Super-PON), a prominent next generation Ethernet PON (NG-EPON) candidate, can suffice this exponentially increasing data rate requirements. However, to appease such requirements, it employs many transceivers and increases the power-consumption of the network. In this work, we focus on reducing the carbon footprint of Super-PON and propose power-efficient dynamic bandwidth and wavelength allocation (DBWA) algorithms, namely best fit bin-packing sleep mode aware (BF-SMA) and updated BF-SMA (UBF-SMA). The proposed algorithms use SMA for bandwidth scheduling and different bin-packing techniques for wavelength allocation. In bin-packing, the number of available wavelengths and their efficient allocation is based on the network load. For restricting the number of available wavelengths, we can switch off the non-essential transceivers at the OLT, which also helps in maximizing the wavelength utilization and increasing the power efficiency. The simulation results show that in comparison to the state-of-the-art DBWA algorithms, the proposed algorithms improve the power efficiency and reduce the average delay of a Super-PON system. Furthermore, we use Jain’s fairness index to validate the fairness of the proposed DBWA algorithms.

Achieving arbitrary throughput-fairness trade-offs in the inter cell interference coordination with fixed transmit power problem

We study the problem of inter cell interference coordination (ICIC) with fixed transmit power in OFDMA-based cellular networks, in which each base station (BS) needs to decide as to which subchannel, if any, to allocate to each of its associated mobile stations (MS) for data transmission. In general, there exists a trade-off between the total throughput (sum of throughputs of all the MSs) and fairness under the allocations found by resource allocation schemes. We introduce the concept of \(\tau -\alpha -\)fairness by modifying the concept of \(\alpha -\)fairness, which was earlier proposed in the context of congestion control for packet-switched networks. The concept of \(\tau -\alpha -\)fairness allows us to achieve arbitrary trade-offs between total throughput and fairness by selecting an appropriate value of \(\alpha\) in \([0,\infty )\). We show that for every \(\alpha \in [0,\infty )\) and every \(\tau > 0\), the problem of finding a \(\tau -\alpha -\)fair allocation is NP-Complete. Further, we show that for every \(\alpha \in [0, \infty )\), there exist thresholds such that if the potential interference levels experienced by each MS on every subchannel are above the threshold values, then the problem can be optimally solved in polynomial time by reducing it to the bipartite graph matching problem. Also, we propose a simple, distributed subchannel allocation algorithm for the ICIC problem, which is flexible, requires a small amount of time to operate, and requires information exchange among only neighboring BSs. We investigate via simulations as to how the algorithm parameters should be selected so as to achieve any desired trade-off between the total throughput and fairness. Finally, we compare the performance of the proposed algorithm with those of the simulated annealing based and opportunistic subchannel allocation algorithms in terms of the total throughput, fairness index and computational complexity.

Investigating gender fairness of recommendation algorithms in the music domain

Although recommender systems (RSs) play a crucial role in our society, previous studies have revealed that the performance of RSs may considerably differ between groups of individuals with different characteristics or from different demographics. In this case, a RS is considered to be unfair when it does not perform equally well for different groups of users. Considering the importance of RSs in the distribution and consumption of musical content worldwide, a careful evaluation of fairness in the context of music RSs is crucial. To this end, we first introduce LFM-2b, a novel large-scale real-world dataset of music listening records, comprising a subset to investigate bias of RSs regarding users’ demographics. We then define a notion of fairness based on the performance gap of a RS between the users with different demographics, and evaluate a variety of collaborative filtering algorithms in terms of accuracy and beyond-accuracy metrics to explore the fairness in the RS results toward a specific gender group. We observe the existence of significant discrepancies (unfairness) between the performance of algorithms across male and female user groups. Based on these discrepancies, we explore to what extent recommender algorithms lead to intensifying the underlying population bias in the final results. We also study the effect of a resampling strategy, commonly used as debiasing method , which yields slight improvements in the fairness measures of various algorithms while maintaining their accuracy and beyond-accuracy performance.

QoS and QoE aware multi objective resource allocation algorithm for cloud gaming

Cloud gaming is an innovative model that congregates video games. The user may have different Quality-of-Experience (QoE), which is a term used to measure a user’s level of satisfaction and enjoyment for a particular service. To guarantee general satisfaction for all users with limited cloud resources, it becomes a major issue in the cloud. This paper leverages a game theory in the cloud gaming model with resource optimization to discover optimal solutions to resolve resource allocation. The Rider-based harmony search algorithm (Rider-based HSA), which is the combination of Rider optimization algorithm (ROA) and Harmony search algorithm (HSA), is proposed for resource allocation to improve the cloud computing system’s efficiency. The fitness function is newly devised considering certain QoE parameters, which involves fairness index, Quantified experience of players (QE), and Mean Opinion Score (MOS). The proposed Rider-based HSA showed better performance compared to Potential game-based optimization algorithm, Proactive resource allocation algorithm, QoE-aware resource allocation algorithm, Distributed algorithm, and Yusen Li et al., with maximal fairness of 0.999, maximal MOS of 0.873, and maximal QE of 1.

DENOUNCER

The use of automated data-driven tools for decision-making has gained popularity in recent years. At the same time, the reported cases of algorithmic bias and discrimination increase as well, which in turn lead to an extensive study of algorithmic fairness. Numerous notions of fairness have been proposed, designed to capture different scenarios. These measures typically refer to a "protected group" in the data, defined using values of some sensitive attributes. Confirming whether a fairness definition holds for a given group is a simple task, but detecting groups that are treated unfairly by the algorithm may be computationally prohibitive as the number of possible groups is combinatorial. We present a method for detecting such groups efficiently for various fairness definitions. Our solution is implemented in a system called DENOUNCER, an interactive system that allows users to explore different fairness measures of a (trained) classifier for a given test data. We propose to demonstrate the usefulness of DENOUNCER using real-life data and illustrate the effectiveness of our method.

Coefficient-Scaling-Based Fair Power Allocation for Multi-User Power-Domain Nonorthogonal Multiple Access Network

Fair power allocation to the users in a power domain nonorthogonal multiple access (PD-NOMA) networks is an essential aspect. In this paper, we have proposed a coefficient-scaling-based fair power allocation scheme, for a single cell multi-user PD-NOMA network. First, we have developed a model for two users with a fixed power allocation scheme, and proposed a modified strong user fair-power allocation scheme. Furthermore, this two-user scheme is extended for an ad-hoc multi-user PD-NOMA network and a coefficient-scaling-based approach for fair power allocation is proposed to address the outage issue. The performance of the proposed schemes is evaluated through bit-error-rate, sum-rate, outage, and fairness index. The simulation results reveal a remarkable increase in fairness performance of the proposed scheme, in comparison with a state-of-the-art method.

Why fairness cannot be automated: Bridging the gap between EU non-discrimination law and AI

In recent years a substantial literature has emerged concerning bias, discrimination, and fairness in artificial intelligence (AI) and machine learning. Connecting this work to existing legal non-discrimination frameworks is essential to create tools and methods that are practically useful across divergent legal regimes. While much work has been undertaken from an American legal perspective, comparatively little has mapped the effects and requirements of EU law. This Article addresses this critical gap between legal, technical, and organisational notions of algorithmic fairness. Through analysis of EU non-discrimination law and jurisprudence of the European Court of Justice (ECJ) and national courts, we identify a critical incompatibility between European notions of discrimination and existing work on algorithmic and automated fairness. A clear gap exists between statistical measures of fairness as embedded in myriad fairness toolkits and governance mechanisms and the context-sensitive, often intuitive and ambiguous discrimination metrics and evidential requirements used by the ECJ; we refer to this approach as “contextual equality.”This Article makes three contributions. First, we review the evidential requirements to bring a claim under EU non-discrimination law. Due to the disparate nature of algorithmic and human discrimination, the EU's current requirements are too contextual, reliant on intuition, and open to judicial interpretation to be automated. Many of the concepts fundamental to bringing a claim, such as the composition of the disadvantaged and advantaged group, the severity and type of harm suffered, and requirements for the relevance and admissibility of evidence, require normative or political choices to be made by the judiciary on a case-by-case basis. We show that automating fairness or non-discrimination in Europe may be impossible because the law, by design, does not provide a static or homogenous framework suited to testing for discrimination in AI systems.Second, we show how the legal protection offered by non-discrimination law is challenged when AI, not humans, discriminate. Humans discriminate due to negative attitudes (e.g. stereotypes, prejudice) and unintentional biases (e.g. organisational practices or internalised stereotypes) which can act as a signal to victims that discrimination has occurred. Equivalent signalling mechanisms and agency do not exist in algorithmic systems. Compared to traditional forms of discrimination, automated discrimination is more abstract and unintuitive, subtle, intangible, and difficult to detect. The increasing use of algorithms disrupts traditional legal remedies and procedures for detection, investigation, prevention, and correction of discrimination which have predominantly relied upon intuition. Consistent assessment procedures that define a common standard for statistical evidence to detect and assess prima facie automated discrimination are urgently needed to support judges, regulators, system controllers and developers, and claimants.Finally, we examine how existing work on fairness in machine learning lines up with procedures for assessing cases under EU non-discrimination law. A ‘gold standard’ for assessment of prima facie discrimination has been advanced by the European Court of Justice but not yet translated into standard assessment procedures for automated discrimination. We propose ‘conditional demographic disparity’ (CDD) as a standard baseline statistical measurement that aligns with the Court's ‘gold standard’. Establishing a standard set of statistical evidence for automated discrimination cases can help ensure consistent procedures for assessment, but not judicial interpretation, of cases involving AI and automated systems. Through this proposal for procedural regularity in the identification and assessment of automated discrimination, we clarify how to build considerations of fairness into automated systems as far as possible while still respecting and enabling the contextual approach to judicial interpretation practiced under EU non-discrimination law.

Dynamic maintenance model for high average-utility pattern mining with deletion operation

The high average-utility itemset mining (HAUIM) was established to provide a fair measure instead of genetic high-utility itemset mining (HUIM) for revealing the satisfied and interesting patterns. In practical applications, the database is dynamically changed when insertion/deletion operations are performed on databases. Several works were designed to handle the insertion process but fewer studies focused on processing the deletion process for knowledge maintenance. In this paper, we then develop a PRE-HAUI-DEL algorithm that utilizes the pre-large concept on HAUIM for handling transaction deletion in the dynamic databases. The pre-large concept is served as the buffer on HAUIM that reduces the number of database scans while the database is updated particularly in transaction deletion. Two upper-bound values are also established here to reduce the unpromising candidates early which can speed up the computational cost. From the experimental results, the designed PRE-HAUI-DEL algorithm is well performed compared to the Apriori-like model in terms of runtime, memory, and scalability in dynamic databases.

Fairness in credit scoring: Assessment, implementation and profit implications

The rise of algorithmic decision-making has spawned much research on fair machine learning (ML). Financial institutions use ML for building risk scorecards that support a range of credit-related decisions. Yet, the literature on fair ML in credit scoring is scarce. The paper makes three contributions. First, we revisit statistical fairness criteria and examine their adequacy for credit scoring. Second, we catalog algorithmic options for incorporating fairness goals in the ML model development pipeline. Last, we empirically compare different fairness processors in a profit-oriented credit scoring context using real-world data. The empirical results substantiate the evaluation of fairness measures, identify suitable options to implement fair credit scoring, and clarify the profit-fairness trade-off in lending decisions. We find that multiple fairness criteria can be approximately satisfied at once and recommend separation as a proper criterion for measuring the fairness of a scorecard. We also find fair in-processors to deliver a good balance between profit and fairness and show that algorithmic discrimination can be reduced to a reasonable level at a relatively low cost. The codes corresponding to the paper are available on GitHub.

Fairness analysis and compensation strategy in the Triangle of Central China driven by water-carbon-ecological footprints.

This study proposes water-carbon-ecological footprints to form footprint family indicators for identifying the ecological compensation and regional development equilibrium in the Triangle of Central China (TOCC). The occupation of natural capital stock and flow consumption can be illustrated using a three-dimensional ecological footprint model, and Gini coefficient is integrated into the evaluation framework for fairness measurement from various aspects. Quantificational ecological compensation standards can be given with considering indicators associated with ecological resource conversion efficiency and willingness to pay. Results reveal that ecological and carbon footprints in the TOCC demonstrate rising trends from 2000 to 2015, while its water footprint presents a fluctuating trend. The majority of average Gini coefficients exceed the threshold value of 0.4 under different footprints, thereby indicating poor overall fairness of regional development. Water footprint in Jingmen, Xiangtan, and Yichun show relatively higher compensation expenses, while Yichang, Zhuzhou, and Fuzhou exhibit higher received compensation values compared with other cities. Carbon footprint in Wuhan, Loudi, and Xinyu indicate high compensation expenses due to their overuse of biological resources. Maximum amounts of compensation expense appear in Nanchang and Wuhan from the perspective of ecological footprint. This study can provide a theoretical reference for sustainable development in the TOCC by performing a comparative analysis with Beijing-Tianjin-Hebei urban agglomeration and developed countries.

The Cloze Procedure as a Measure of Program Understanding.(Dept.E)

The process of program understanding plays an important role in testing, and maintenance phases of software life cycle. There have been numerous investigations of the influence of various aspects of a program on program understanding, using many different measures of understanding. Some of the different measures include time to find a bug, a comprehension quiz, ability to reporduce a functionally equivalent program without notes, time to perform modification. All of these measures have some limitations. A new measure called CLOZE procedure, which has less limitations as well as other advantages such as ease of administration and grading, is investigated in this paper . The paper reports on a controlled experiment that compared CLOZE procedure and the time to find a bug as measuring program understanding. The experiment results indicated that the CLOZE procedure measure is a fair measure for program understanding. It may best be used in educational environments where computerizing test construction, as well as grading, is highly appreciated.

Energy-efficient Dynamic Mobile Sink Path Planning for Data Acquisition for Wireless Sensor Networks

Wireless sensor networks (WSNs) are widely used in various applications such as defense, forest fire, healthcare,structural health monitoring, etc., because of itsflexibility, low cost and tiny. In WSNs, the sensor nodes are scattered over the target area to acquire the data from the environment and transmit it to the base station via single or multi-hop communication. Due to the sensor nodes' constrained battery, the sensor nodes near the base station are more involved in data transmissions. These relay nodes drain more energy and die soon, leading to a hotspot/energy-hole problem. Several algorithms have been proposed in the literature to address the hotspot problem using the mobile sink. However, most of the existing approaches are highlycomputational and also provide a static solutiononly. In this context, we proposed an energy-efficient dynamic mobile sink path construction with low computational complexity for data acquisition in WSNs. We use the minimum spanning tree-based clustering for selecting the data collection points and a computational geometry-based method to identify the visiting order of the data collection points by the mobile sink. Our proposed work is better than the existing approaches in terms of average energy consumption, network lifetime, fairness index, buffer utilization, etc.

Verification of a new metric for data flow continuity and sustainability optimization

SummaryThis article presents a verification process of a new metric designed for transmission continuity optimization. Unlike the commonly used metrics, it focuses solely on flow continuity and sustainability, disregarding the maximization of throughput or minimization of latency. We base our metric on inter‐packet gaps analysis and ordering of packets from multiple communication flows into a final flow. We also compare our metric with the commonly used one based on Jain's fairness index in cases when optimization of communication flow sustainability is the priority. We present the evaluation of results and analyze the data processing complexity. Moreover, we discuss the weaknesses and strengths of different approaches to communication flow optimization.

Non-Parametric Shape Design of Free-Form Shells Using Fairness Measures and Discrete Differential Geometry

A non-parametric approach is proposed for shape design of free-form shells discretized into triangular mesh. The discretized forms of curvatures are used for computing the fairness measures of the surface. The measures are defined as the area of the offset surface and the generalized form of the Gauss map. Gaussian curvature and mean curvature are computed using the angle defect and the cotangent formula, respectively, defined in the field of discrete differential geometry. Optimization problems are formulated for minimizing various fairness measures for shells with specified boundary conditions. A piecewise developable surface can be obtained without a priori assignment of the internal boundary. Effectiveness of the proposed method for generating various surface shapes is demonstrated in the numerical examples.

Energy Efficient Downlink Resource Allocation in Cellular IoT Supported H-CRANs

The cloud computing supported heterogeneous cloud radio access network (H-CRAN) is one of the promising solutions to support cellular IoT devices with the legacy cellular systems. However, the dense deployment of small cells with fractional frequency reuse in orthogonal frequency division multiple access (OFDMA) based H-CRANs increases intra- and inter-cell interference, turning the resource allocation into a more challenging problem. In general, the macro cell users are considered as the legacy users, whereas the cellular IoT devices and small cell users share the macro cell users' resource blocks in an underlaid approach. In this paper, we investigate an underlaid approach of resource allocation for small and macro cell users to improve the energy efficiency (EE) in H-CRANs. The solution approaches are derived with the Dinkelbach, Lagrange multiplier and Alternating Direction Method of Multipliers (ADMM) methods by considering the maximum power, resource block allocation, fronthaul capacity and quality of service (QoS) constraints of macro cell users. A two-step energy efficient underlaid cellular IoT (UC-IoT) supported H-CRAN method is proposed and evaluated with the overlaid cellular IoT (OC-IoT) supported H-CRAN and underlaid H-CRAN without cellular IoT devices. The proposed method is evaluated in terms of energy efficiency and the Jain's fairness index, considering the effect of number of cellular IoT density in each small cell of the H-CRAN. The simulation results demonstrate the effectiveness of the proposed approach compared to earlier approaches.