Artificiell intelligens har under senare år tagit en allt större roll inom företagsvärlden. Organisationer kan utnyttja AI till att effektivisera rekryteringsprocessen vilket sparar tid och pengar. Studier visar att träning av AI kan leda till diskriminering. Detta är ofta en sidoeffekt till hur AI har tränats, exempelvis genom data innehållandes bias, eller att programmerarens egna fördomar blir del av AI:s beslutsunderlag. Syfte: Syftet med den här studien är att genom en systematisk litteraturstudie skapa en inblick i de konsekvenser som användandet av AI kan ha på rekryteringsprocessen. Rapporten ämnar även att undersöka hur användandet av AI kan kopplas till diskriminering av kandidater. Avslutningsvis önskar studien att ge rekommendationer till hur AI i fortsättningen kan användas i rekryteringsprocessen utan att åsamka skada på organisationer eller individer. Resultat: Resultatet tyder på att rekryterare och kandidater upplever AI som bias. Resultaten visar även vikten av transparens och det fortsatta behovet av människor i rekryteringsprocessen, för att öka förtroendet och undvika diskriminering och partiskt beslutsfattande från AI. Dessutom visar studien att kandidater uppfattar att AI inte kan identifiera unika drag och personliga ansträngningar från kandidater, och därför anser tekniken som orättvis. Den här studien observerar en polariserad syn på AI som å ena sidan mer partisk än människor, och å andra sidan mindre partisk än människor. Detta förmår författarna till den här rapporten att diskutera komplexiteten i ämnet, och att AI liksom människor, kan orsaka partiskhet. Slutsats: Kandidater och rekryterare anser att AI i rekrytering är fördomsfullt. Uppfattningen påverkas av individuella aspekter. AI i rekrytering kan leda till diskriminering av ålder, kön och etnicitet, diskrimineringen kan ofta härledas till hur AI har tränats. AI kan användas i de delar av rekrytering som historiskt inneburit lite kontakt mellan organisation och kandidat, tillexempel CV granskning. Mänsklig närvaro behövs för att öka förtroende hos kandidater. / Artificial intelligence has during recent years taken a significant place in the business world. Organisations benefit from AI by streamlining the recruitment process and thus saving money by reducing cost and time. Studies show how training of AI can cause discrimination. This is often a side effect of how the AI has been trained, using data containing bias, or due to the programmers personal bias being incorporated in the AI decision making. Purpose: The purpose of this study is to create insight into the consequences that the use of AI can have on the recruitment process. Furthermore, this study aims to examine how AI can be linked to discrimination against candidates. Lastly, this study wishes to give recommendations on how AI can be used for recruitment in the future, without causing harm to individuals or organisations. This study has analysed contemporary management research concerning the area using a systematic literature review. Results: The literature suggests that AI is perceived by recruiters and candidates as biassed. The results show the importance of transparency and the continued need for humans in the process, to enhance trust and avoid discrimination and biassed decision making from AI. Furthermore, this study shows that candidates perceive that AI is unable to identify uniqueness and personal efforts and thus believe the technology to be unfair. This study observes a polarised view of AI as more biassed than humans on one hand, and less biassed than humans on the other. This induces the authors of this paper to discuss the complexity of the matter and that AI, like humans, can cause bias. Conclusion: Candidates and recruiters believe that AI in recruitment is biassed. The perception is influenced by individual aspects. AI in recruitment can lead to discrimination of age, gender and ethnicity, the discrimination can often be traced to how the AI has been trained. AI can be used in those parts of recruitment that historically involved little contact between organisation and candidate, such as CV scanning. Human presence is needed to increase trust in candidates.

We conducted experiments on norm focusing. The tests were carried out with two versions of dictator games: in one version of the game, the dictator had to allocate a gain of e10, while in the other version, a loss of e−10 needs to be allocated. In a first treatment, we focused subjects on the average giving in similar previous dictator games. The second treatment focused subjects on the behaviour of what a self-interested actor should do. In total, N = 550 participants took part in our experiments. We found (1) a significant difference in giving behaviour between gain and loss treatments, with subjects being moderately more self-interested in the loss domain, (2) a significant effect of focusing subjects on the average behaviour of others, but (3) no effect of focusing subjects on the behaviour of self-interested actors.

The last decade has brought with it a wave of innovative technology, shifting the channels through which creative content is created, consumed, and categorized. And, as our interactions with creative multimedia content shift towards online platforms, the sheer quantity of content on these platforms has necessitated the integration of algorithmic guidance in the discovery of these spaces. In this way, the recommendation algorithms that guide users' interactions with various art forms have been cast into the role of gatekeepers and begun to play an increasingly influential role in shaping the creation of artistic content. The work laid out in the following chapters fuses three major areas of research: graph representation learning, music information retrieval, and fairness as applied to the task of music recommendation. In recent years, graph neural networks (GNNs), a powerful new architecture which enables deep learning approaches to be applied to graph or network structures, have proven incredibly influential in the music recommendation domain. In tandem with the striking performance gains that GNNs are able to achieve, many of these systems, have been shown to be strongly influenced by the degree, or number of outgoing edges, of individual nodes. More concretely, recent works have uncovered disparities in the qualities of representations learned by state of the art GNNs between nodes which are strongly and weakly connected. Translating these findings to the sphere of recommender systems, where nodes and edges are used to represent the interactions between users and various items, these disparities in representation that are contingent upon a node's connectivity can be seen as a form of popularity bias. And, indeed, within the broader recommendation community, popularity bias has long been considered an open problem, in which recommender systems begin to favor mainstream content over, potentially more relevant, but niche or novel items. If left unchecked these algorithmic nudged towards previously popular content can create, intensify, and enforce negative cycles that perpetuate disparities in representation on both the user and the creator ends of the content consumption pipeline. Particularly in the recommendation of creative (e.g. musical) content, the downstream effects in these disparities of visibility can have genuine economic consequences for artists from under-represented communities. Thus, the problem of popularity bias is something that must be addressed from both a technical and societal perspective. And, as the influence of recommender systems continues to spread, the effects of this phenomenon only become more spurious, as they begin to have critical downstream effects that shape the larger ecosystems in which art is created. Thus, the broad focus of thesis is the mitigation of popularity bias in music recommendation. In order to tailor our exploration of this issue to the graph domain, we begin by formalizing the relationship between degree fairness and popularity bias. In doing so, we concretely define the notion of popularity, grounding it in the structural principles of an interaction network, and enabling us to design objectives that can mitigate the effects of popularity on representation learning. In our first work, we focus on understanding the effects of sampling on degree fairness in uni-partite graphs. The purpose of this work is to lay the foundation for the graph neural network model which will underlie our music recommender system. We then build off this first work by extending the initial fairness framework to be compatible with bi-partite graphs and applying it to the music domain. The motivation of this work is rooted in the notion of discovery, or the idea that users engage with algorithmic curation in order to find content that is both novel and relevant to their artistic tastes. We present the intrinsic relationship between discovery objectives and the presence of popularity bias, explaining that the presence of popularity bias can blind a system to the musical qualities that underpin the underlying needs of music listening. As we will explain in later sections, one of the key elements of this work is our ability to ground our fairness notion in the musical domain. Thus, we propose a domain-aware, individual fairness-based approach which addresses popularity bias in graph neural network (GNNs) based recommender systems. In order to facilitate this domain awareness, we perform extensive dataset augmentation, taking two state of the art music recommendation datasets and augmenting them with rich multi-modal node-level features. Finally, we ground our evaluation in the cold start setting, showing the importance of inductive methodologies in the music space. / La dernière décennie a apporté avec elle une vague de technologies innovantes, modifiant la manière dont le contenu créatif est créé, consommé et catégorisé. Et, à mesure que nos interactions avec les contenus multimédias créatifs se déplacent vers les plateformes en ligne, la quantité de contenu sur ces plateformes a nécessité l’intégration d’un guidage algorithmique dans la découverte de ces espaces. De cette façon, les algorithmes de recommandation qui guident les interactions des utilisateurs avec diverses formes d’art ont été jetés dans le rôle de gardiens et ont commencé à jouer un rôle de plus en plus influent dans l’élaboration de la création de contenu artistique. Le travail présenté dans les chapitres suivants fusionne trois grands domaines de recherche : l’apprentissage de la représentation graphique, la recherche d’informations musicales et l’équité appliquée à la tâche de recommandation musicale. Alors que l’influence des systèmes de recommandation continue de s’étendre et de s’intensifier, il est crucial de prendre en compte les effets en aval que les choix de conception peuvent avoir sur l’écosystème plus large de la création artistique. Ces dernières années, l’intégration des réseaux sociaux dans la tâche de recommandation musicale a donné naissance aux réseaux neuronaux de graphes (GNN), une nouvelle architecture capable de faire des prédictions sur les structures de graphes. Parallèlement aux gains miraculeux que les GNN sont capables de réaliser, bon nombre de ces systèmes peuvent également être la proie de biais de popularité, les forçant à privilégier le contenu grand public par rapport à des éléments potentiellement plus pertinents, mais de niche ou nouveaux. S’il n’est pas maîtrisé, ce cycle négatif peut perpétuer les disparités de représentation entre la musique d’artistes, de genres ou de populations minoritaires. Et, ce faisant, les disparités dans la visibilité des éléments peuvent entraîner des problèmes à la fois du point de vue des performances et de la société. L’objectif de la thèse est l’atténuation du biais de popularité. Premièrement, le travail formalise les liens entre l’équité individuelle et la présence d’un biais de popularité parmi les contenus créatifs. Ensuite, nous étendons un cadre d’équité individuelle, en l’appliquant au domaine de la recommandation musicale. Le coeur de cette thèse s’articule autour de la proposition d’une approche basée sur l’équité individuelle et sensible au domaine qui traite le biais de popularité dans les systèmes de recommandation basés sur les réseaux de 5 neurones graphiques (GNN). L’un des éléments clés de ce travail est notre capacité à ancrer notre notion d’équité dans le domaine musical. Afin de faciliter cette prise de conscience du domaine, nous effectuons une augmentation étendue des ensembles de données, en prenant deux ensembles de données de recommandation musicale à la pointe de la technologie et en les augmentant avec de riches fonctionnalités multimodales au niveau des noeuds. Enfin, nous fondons notre évaluation sur le démarrage à froid, montrant l’importance des méthodologies inductives dans l’espace musical.

No description available.

No description available.

<p dir="ltr">Despite the recent advancement and exponential growth in the utility of deep learning models across various fields and tasks, we are confronted with emerging challenges. Among them, one prevalent issue is the biases inherent in deep models, which often mimic stereotypical or subjective behavior observed in data, potentially resulting in negative societal impact or disadvantaging certain subpopulations based on race, gender, etc. This dissertation addresses the critical problem of fairness and bias in machine learning from diverse perspectives, encompassing both data biases and model biases.</p><p dir="ltr">First, we study the multifaceted nature of data biases to comprehensively address the challenges. Specifically, the proposed approaches include the development of a generative model for balancing data distribution with counterfactual samples to address data skewness. In addition, we introduce a novel feature selection method aimed at eliminating sensitive-relevant features that could potentially convey sensitive information, e.g., race, considering the interrelationship between features. Moreover, we present a scalable thresholding method to appropriately binarize model outputs or regression data considering fairness constraints for fairer decision-making, extending fairness beyond categorical data.</p><p dir="ltr">However, addressing fairness problem solely by correcting data bias often encounters several challenges. Particularly, establishing fairness-curated data demands substantial resources and may be restricted by regal constraints, while explicitly identifying the biases is non-trivial due to their intertwined nature. Further, it is important to recognize that models may interpret data differently by their architectures or downstream tasks. In response, we propose a line of methods to address model bias, on top of addressing the data bias mentioned above, by learning fair latent representations. These methods include fair disentanglement learning, which projects latent subspace independent of sensitive information by employing conditional mutual information, and a debiased contrastive learning method for fair self-supervised learning without sensitive attribute annotations. Lastly, we introduce a novel approach to debias the multimodal embedding of pretrained vision-language models (VLMs) without requiring downstream annotated datasets, retraining, or fine-tuning of the large model considering the constrained resource of research labs.</p>

Recently, there has been a massive growth in the number of mobile users and their traffic. The data traffic volume almost doubles every year. Mobile users are currently running multiple applications that require higher bandwidth which makes users so limited to the service providers' resources. Increasing the utilization of the existing spectrum can significantly improve network capacity, data rates and user experience. Spectrum sharing enables wireless systems to harvest under-utilized swathes of spectrum, which would vastly increase the efficiency of spectrum usage. Making more spectrum available can provide significant gain in mobile broadband capacity only if those resources can be aggregated efficiently with the existing commercial mobile system resources. Carrier aggregation (CA) is one of the most distinct features of 4G systems including Long Term Evolution Advanced (LTE-Advanced). In this dissertation, a resource allocation with carrier aggregation framework is proposed to allocate multiple carriers resources optimally among users with elastic and inelastic traffic in cellular networks. We use utility proportional fairness allocation policy, where the fairness among users is in utility percentage of the application running on the user equipment (UE). A resource allocation (RA) with CA is proposed to allocate single or multiple carriers resources optimally among users subscribing for mobile services. Each user is guaranteed a minimum quality of service (QoS) that varies based on the user's application type. In addition, a resource allocation with user discrimination framework is proposed to allocate single or multiple carriers resources among users running multiple applications. Furthermore, an application-aware resource block (RB) scheduling with CA is proposed to assign RBs of multiple component carriers to users' applications based on a utility proportional fairness scheduling policy. We believe that secure spectrum auctions can revolutionize the spectrum utilization of cellular networks and satisfy the ever increasing demand for resources. Therefore, a framework for multi-tier dynamic spectrum sharing system is proposed to provide an efficient sharing of spectrum with commercial wireless system providers (WSPs) with an emphasis on federal spectrum sharing. The proposed spectrum sharing system (SSS) provides an efficient usage of spectrum resources, manages intra-WSP and inter-WSP interference and provides essential level of security, privacy, and obfuscation to enable the most efficient and reliable usage of the shared spectrum. It features an intermediate spectrum auctioneer responsible for allocating resources to commercial WSPs' base stations (BS)s by running secure spectrum auctions. In order to insure truthfulness in the proposed spectrum auction, an optimal bidding mechanism is proposed to enable BSs (bidders) to determine their true bidding values. We also present a resource allocation based on CA approach to determine the BS's optimal aggregated rate allocated to each UE from both the BS's permanent resources and winning auctioned spectrum resources. / Ph. D.

Yes / Recent decades have witnessed a growing number of states around the world relying on border control measures, such as immigration detention, to govern human mobility and control the movements of those classified as “unauthorised non-citizens.” In response to this, an increasing number of scholars from several disciplines, including psychologists, have begun to examine this phenomenon. In spite of the widespread concerns raised, few studies have been conducted inside immigration detention sites, primarily due to difficulties in gaining access. This body of research becomes even scanter when it comes to the experiences of detained women. This study is the first of its kind to have surveyed 93 women confined in an Italian immigration detention facility. A partial mediation model with latent variables was tested through partial least structural equation modelling (PLS-SEM). The findings revealed the negative impact that unfair immigration procedures have on detained women’s human dignity, which in turn negatively affects their self-rated physical and mental health. Overall, our study sheds light on the dehumanisation and damage to human dignity that immigration detention entails, as well as its negative impact on the health of those affected. This evidence reinforces the image of these institutions as sites of persistent injustice, while stressing the need to envision alternative justice-oriented forms to address human mobility. / FE’s doctoral research, on which this article relies, was supported by the Portuguese Foundation for Science and Technology (SFRH/BD/87854/2012), and her subsequent work was supported by the Portuguese Foundation for Science and Technology (grant number: CEECIND/00924/2018/CP1541/CT0004).

The thesis analyses the concept ‘fairness’ in consumer contracts regulated by the Consumer Protection Act 68 of 2008, mainly from the perspective of a freedom and fairness orientation. It discusses the evolution of ‘fairness’ as background to a more detailed discussion of the classification of fairness into substantive and procedural fairness. The thesis examines dimensions of fairness, factors which play a role in the determination of fairness, and fairness- oriented approaches in an attempt to formulate a framework for fairness in consumer contracts. The main aspects that should be taken into account to justify a finding of fairness, or to determine whether a contract is fair, are identified. This analysis addresses, too, the extent to which the fairness provisions of the Consumer Protection Act are appropriate (with reference to the law of South Africa, Europe, and England). / Mercantile Law / LL.D.

The dissertation examines the applicability of procedural fairness to actions by members of the South African National Defence Forces (SANDF). The research focuses on and uses the South African Defence Force Union v The Minister of South African National Defence Force (SANDU 2010 judgment) to illustrate how procedural fairness should find application in the SANDF, given the sui generis nature of the defence forces. This judgment presented an opportunity to investigate whether the legislative framework that is available in the SANDF is adequate to protect the right to procedural fairness of the members of the SANDF encapsulated in section 33 of the Constitution, 1996. The dissertation examines the relevant sections of the Defence Act, Military Discipline Supplementary Measures Act, Labour Relations Act (LRA), and the Promotion of Administrative Justice Act (PAJA) read with sections 23 and 33 of the Constitution to determine whether there is a gap that exists in so far as the protection of the right to procedural fairness of members of the defence forces is concerned. It also examines the Military Discipline Code and the rules and regulations of the Defence Forces. The analysis of the SANDU 2010 judgment demonstrates that PAJA could find application in dismissal or employment related disputes within the SANDF. The scenario that is evidenced from the analysis of the defence force legislative framework is that the legislative framework that is available within the SANDF is inadequate to protect and deal with disputes which arise from allegations of infringement of the right to procedural fairness. This scenario is compounded by the fact that the LRA which is the empowering legislation that was promulgated to give effect to the right to section 23 of the Constitution and to deal with dismissal and employment related disputes, does not apply to members of the SANDF. / Public, Constitutional, and International Law / LL. M.