No description available.

This study aimed to explore the implementation of fairness in intelligent agents to enhance their interactions in our social space. Two distinct investigations, an experiment, and a focus group, were conducted to examine the impact of unfair treatment by non-anthropomorphic and anthropomorphic agents, where we sought to answer the research question: How does experiencing unfair treatment from agents with different appearances influence individuals' perceptions, satisfaction, and trust? The experiment encompassed four experimental conditions combining fair and unfair behaviours with agents displaying human-like or non-human-like appearances. User enactment, Experience prototyping, and the Wizard of Oz technique were employed during the experiment. The focus group aimed to delve into the concept of fairness and its relevance to agents in greater detail. In summary, the study's findings indicate that fairness is a significantly important consideration in agent design. However, the complexity of designing a fair agent proves challenging, due to the subjective and contextual nature where it entangles with various factors. / Toward socially competent AI: Designing multi-user interaction with embodied intelligent agents to support politeness and fairness (SCAI)

Le 21 septembre 2021, l’Assemblée nationale du Québec a adopté le projet de loi 64 afin de moderniser son régime de protection des renseignements personnels. S’inspirant du Règlement Général sur la Protection des Données européen, ce projet de loi renforce substantiellement les obligations des entreprises privées et des organismes publics à l’égard des renseignements personnels des Québécois. Ce projet de loi assure également le respect de certains principes juridiques applicables aux systèmes d’intelligence artificielle. Or, dans le cadre de ce mémoire, nous démontrons que des tensions existent entre quatre de ces principes. Ces principes sont : le principe d’explicabilité, le principe d’exactitude, le principe de sécurité ainsi que le principe d’équité et de non-discrimination. En effet, il est souvent difficile et parfois impossible d’assurer un respect conjoint de ces quatre principes. La présente étude se divise en trois chapitres. Le premier explore les quatre principes pour ensuite identifier les obligations légales québécoises qui permettent d’en assurer le respect. Le second expose les tensions entre ces principes. Le dernier propose une solution permettant aux entreprises et aux organismes publics québécois de réaliser les arbitrages nécessaires entre ces principes tout en respectant la Loi. / On September 21, 2021, the Quebec legislative passed Bill 64 to modernize its privacy regime. Inspired by the European General Data Protection Regulation, this bill strengthens the obligations of private companies and public bodies with respect to personal data. This bill also provides obligations protecting normative principles applicable to artificial intelligence systems. In this paper, we show that four of these principles exist in a state of tension. These principles are : explicability, accuracy, security and fairness and non-discrimination. Indeed, it is often difficult and sometimes impossible to ensure that these principles are respected together. This study is divided into three parts. The first part defines the four principles to then identifies how these principles are translated into Quebec law. The second part sets out the tensions between these principles. The last part provides a solution that would allow Quebec businesses and public bodies to make the necessary trade-offs between these principles in a matter that complies with their legal obligations.

Artificial intelligence, AI, is widely acknowledged to have atransformative impact on various industries. However, thistechnology is not without its limitations. One such limitationis the potential reinforcement of human biases withinmachine learning systems. After all, these systems rely ondata generated by humans. To address this issue, theEuropean Union, EU, are implementing regulationsgoverning the development of AI systems, not only topromote ethical decision-making but also to curb marketoligopolies. Achieving fair decision-making relies on highquality data. The performance of a model is thussynonymous with high-quality data, encompassing breadth,accurate annotation, and relevance. Previous researchhighlights the lack of processes and methods guiding theeffort to ensure high-quality training data. In response, thisstudy aims to investigate the limitations and opportunitiesassociated with claims of data quality within the domain ofdata collection research. To achieve this, a research questionis posed: What factors constrain and enable the creation of ahigh-quality dataset in the context of AI fairness? The studyemploys a method of semistructured interviews withindustry experts, allowing them to describe their personalexperiences and the challenges they have encountered. Thestudy reveals multiple factors that restrict the ability tocreate a high-quality dataset and, ultimately, a fair decisionmaking system. The study also reveals a few opportunities inrelation to high quality data, which methods associated withthe research landscape provides. / Att artificiell intelligens är något som kommer vända uppoch ned på många branscher är något som många experter äröverens om. Men denna teknik är inte helt befriad frånbegränsningar. En av dessa begränsningar är att ett systemsom använder maskininlärning potentiellt kan förstärka defördomar vi människor besitter. Tekniken grundar sig trotsallt i data, data som skapas av oss människor. EU harbestämt sig för att tackla denna problematik genom att införaregler gällande huruvida system som tillämpar AI skallutvecklas. Både för att gynna det etiska beslutsfattandet menockså för att hämma oligopol på marknaden. För att uppnåett så rättvist beslutsfattande som möjligt krävs det data avhög kvalitet. En modells prestanda är således synonymt meddata av hög kvalitet, där bredd, korrekt annotering ochrelevans är betydande. Tidigare forskning pekar påavsaknaden av processer och metoder för att vägleda arbetetmed att säkerställa högkvalitativa träningsdata. Som svar pådetta syftar denna studie till att undersöka vilkabegränsningar och möjligheter som gör anspråk pådatakvalitet i delar av forskningsområdet Data collection.Detta görs genom att besvara forskningsfrågan: Vilkafaktorer begränsar och möjliggör skapandet av etthögkvalitativt dataset i kontexten rättvis AI? Metoden somtillämpas i studien för att besvara ovanstående ärsemistrukturerade intervjuer där yrkesverksamma experterfår beskriva sina personliga upplevelser gällande vilkautmaningar de har ställts inför. Studien resulterar i ett antalfaktorer som begränsar förutsättningarna för att skapa etthögkvalitativt dataset och i slutändan ett rättvistbeslutsfattande system. Studien resulterar även i att peka påett antal möjligheter i relation till högkvalitativa data, sommetoder associerade med forskningslandskapet besitter

Le besoin grandissant pour une planification urbaine plus durable et pour des interventions publiques visant à l'amélioration du bien-être collectif, ont grandement contribué à un engouement pour les espaces verts. Les parcs sont reconnus pour leur impact positif en zone urbaine dense, et nous sommes intéressés par l'application des concepts théoriques du domaine de la recherche opérationnelle pour assister les décideurs publics afin d'améliorer l'accessibilité, la distribution et la conception des parcs. Étant donné le contexte, nous sommes particulièrement motivés par le concept d'équité, et étudions le comportement des usagers des parcs à l'aide d'un modèle d'interaction spatiale, tel qu'appliqué dans les problèmes d'emplacement d'installations dans un marché compétitif. Dans cette recherche, nous présentons un modèle d'emplacement d'installations à deux étapes pouvant être adapté pour assister les décideurs publics à l'échelle de la ville. Nous étudions spécifiquement l'application aux espaces verts urbains, mais soulignons que des extensions du modèle peuvent permettre d'aborder d'autres problèmes d'emplacements d'installations sujets à des enjeux d'équité. La première étape de notre problème d'optimisation a pour but d'évaluer l'allocation la plus équitable du budget de la ville aux arrondissements, basé sur une somme du budget pondérée par des facteurs d'équité. Dans la deuxième étape du modèle, nous cherchons l'emplacement et la conception optimale des parcs, et l'objectif consiste à maximiser la probabilité totale que les individus visitent les parcs. Étant donné la non-linéarité de la fonction objective, nous appliquons une méthode de linéarisation et obtenons un modèle de programmation linéaire mixte en nombres entiers, pouvant être résolu avec des solveurs standards. Nous introduisons aussi une méthode de regroupement pour réduire la taille du problème, et ainsi trouver des solutions quasi optimales dans un délai raisonnable. Le modèle est testé à l'aide de l'étude de cas de la ville de Montréal, Canada, et nous présentons une analyse comparative des résultats afin de justifier la performance de notre modèle. / The recent promotion of sustainable urban planning combined with a growing need for public interventions to improve well-being and health in dense urban areas have led to an increased collective interest for green spaces. Parks have proven a wide range of benefits in urban areas, and we are interested in the application of theoretical concepts from the field of Operations Research to assist decision-makers to improve parks' accessibility, distribution and design. Given the context of public decision-making, we are particularly concerned with the concept of fairness, and are focused on an advanced assessment of users' behavior using a spatial interaction model (SIM) as in competitive facility locations' frameworks. In this research, we present a two-stage fair facility location and design (2SFFLD) model, which serves as a template model to assist public decision-makers at the city-level for the urban green spaces (UGSs) planning. We study the application of the 2SFFLD model to UGSs, but emphasize the potential extension to other applications to location problems concerned with fairness and equity. The first-stage of the optimization problem is about the optimal budget allocation based on a total fair-weighted budget formula. The second-stage seeks the optimal location and design of parks, and the objective consists of maximizing the total expected probability of individuals visiting parks. Given the non-linearity of the objective function, we apply a ``Method-based Linearization'' and obtain a mixed-integer linear program that can be solved with standard solvers. We further introduce a clustering method to reduce the size of the problem and determine a close to optimal solution within reasonable time constraints. The model is tested using the case study of the city of Montreal, Canada, and comparative results are discussed in detail to justify the performance of the model.

<p>Mental health significantly impacts issues like gun violence, school shootings, and suicide. There is a strong connection between mental health and emotional states. By monitoring emotional changes over time, we can identify triggering events, detect early signs of instability, and take preventive measures. This thesis focuses on the development of a generalized and modular system for human emotion recognition and explanation based on visual information. The aim is to address the challenges of effectively utilizing different cues (modalities) available in the data for a reliable and trustworthy emotion recognition system. Our face is one of the most important medium through which we can express our emotion. Therefore We first propose SAFER, A novel facial emotion recognition system with background and place features. We provide a detailed evaluation framework to prove the high accuracy and generalizability. However, relying solely on facial expressions for emotion recognition can be unreliable, as faces can be covered or deceptive.  To enhance the system's reliability, we introduce EMERSK, a multimodal emotion recognition system that integrates various modalities, including facial expressions, posture, gait, and scene background, in a flexible and modular manner. It employs convolutional neural networks (CNNs), Long Short-term Memory (LSTM), and denoising auto-encoders to extract features from facial images, posture, gait, and scene background. In addition to multimodal feature fusion, the system utilizes situational knowledge derived from place type and adjective-noun pairs (ANP) extracted from the scene, as well as the spatio-temporal average distribution of emotions, to generate comprehensive explanations for the recognition outcomes. Extensive experiments on different benchmark datasets demonstrate the superiority of our approach over existing state-of-the-art methods. The system achieves improved performance in accurately recognizing and explaining human emotions. Moreover, we investigate the impact of novelty, such as face masks during the Covid-19 pandemic, on the emotion recognition. The study critically examines the limitations of mainstream facial expression datasets and proposes a novel dataset specifically tailored for facial emotion recognition with masked subjects. Additionally, we propose a continuous learning-based approach that incorporates a novelty detector working in parallel with the classifier to detect and properly handle instances of novelty. This approach ensures robustness and adaptability in the automatic emotion recognition task, even in the presence of novel factors such as face masks. This thesis contributes to the field of automatic emotion recognition by providing a generalized and modular approach that effectively combines multiple modalities, ensuring reliable and highly accurate recognition. Moreover, it generates situational knowledge that is valuable for mission-critical applications and provides comprehensive explanations of the output. The findings and insights from this research have the potential to enhance the understanding and utilization of multimodal emotion recognition systems in various real-world applications.</p> <p><br></p>

Denna rapport behandlar prestandaproblem i den nyligen standardiserade Mesh kommunikationsstandarden (IEEE 802.11s). I denna rapport, undersöker och förbättra vi ett förhållande som resulterar i reduktion av genomströmningen i en kedja av noder topologi. IEEE802.11s är mycket lovande med många fördelar för både IoT-systemen och trådlösa nätverk i båda hemmet och arbete.Vi arbetar med frågan om orättvisa när CSMA/CA tillämpas, vilket orsakar genomströmningsreduktion på grund av paketförluster och indikerar svältning. Vi analyserar konsekvenserna av Collision Avoidance (CA) mekanism och föreslår en ersättning för CA som är både rättvist och samtidigt kan upprätthålla undvikande av kollisioner. Vi implementera detta i en simulator och resultatet visar på betydligt högre end-to-end-genomströmning än standard CSMA/CA och inga paketförluster på grund av buffertspill. / This thesis rapport deals with the performance issues of the newly standardized Wireless mesh protocol (IEEE 802.11s). In this thesis, we work on improving the conditions that results in throughput degradation in a chain of nodes topology. The mesh standard is very promising with many advantages for both IoT systems and home wireless networks.We work on the issue of unfairness when CSMA/CA is applied, which causes throughput degradation due to packet loss and indicates starvation. We analyze the implication of the Collision Avoidance (CA) mechanism and propose a replacement for the CA that is both fair and able to maintain collision avoidance. We implement this in a simulator and the result shows significantly higher end-to-end throughput compared to the original CSMA/CA and no packet loss due to buffer overflow.

Equal and fair grading is crucial for the grading system to be perceived as legitimate by society and for the selection to higher education to be legally secure for the student. In a grading system where the teacher has a high degree of autonomy in the grading process, demands are placed on the teacher's integrity and professionalism. This systematic literature study examines which assessment and grading practices can strengthen equal and fair grading, and which difficulties may arise in the tension between the student's legal security and the teacher's professionalism. The study is based on an organizational justice theory with three different perspectives: distributive justice, procedural justice, and interpersonal justice. The main result of the study is that relational and caring assessment and grading practices, a collective interpretation of the grading criteria at national level, and high validity and reliability in the assessment method are strengthening for equal and fair grading. In addition, difficulties arise regarding teachers' assessment and grading practices when the teacher's role as grader is questioned, and the grading process is influenced by internal or external factors. This has consequences for the student's legal security since the teacher's professional judgment stands as the sole guardian of maintaining it. The conclusion is that grading systems that aspire to be equal and fair need principles for how the balance between individual and impartial assessments should be applied.

Facial recognition technology has become a ubiquitous tool in security and personal identification. However, the rise of this technology has been accompanied by concerns over inherent biases, particularly regarding ethnic and gender. This thesis examines the extent of these biases by focusing on the influence of dataset imbalances in facial recognition algorithms. We employ a structured methodological approach that integrates AI-generated images to enhance dataset diversity, with the intent to balance representation across ethnics and genders. Using the ResNet and Vgg model, we conducted a series of controlled experiments that compare the performance impacts of balanced versus imbalanced datasets. Our analysis includes the use of confusion matrices and accuracy, precision, recall and F1-score metrics to critically assess the model’s performance. The results demonstrate how tailored augmentation of training datasets can mitigate bias, leading to more equitable outcomes in facial recognition technology. We present our findings with the aim of contributing to the ongoing dialogue regarding AI fairness and propose a framework for future research in the field.

Verschiedene Schulleistungsstudien stellten für Kinder mit nicht-deutscher Familiensprache bereits in der Grundschule substanzielle Disparitäten im Bereich Mathematik fest. Diese Disparitäten führten zu der Frage, ob die verwendeten Testverfahren zu hohe sprachliche Hürden für Kinder mit nicht-deutscher Familiensprache aufweisen und daher nicht ausreichend in der Lage sind, die Kompetenzen dieser Gruppe valide zu erfassen. In dieser kumulativen Arbeit wurde geprüft, inwiefern die sprachliche Komplexität von Mathematikaufgaben in der Grundschule einen benachteiligenden Einfluss auf die Erfassung der Mathematikleistung von Kindern mit nicht-deutscher Familiensprache darstellt. Zunächst wurde geprüft, ob die in nationalen Schulleistungsstudien verwendeten Aufgaben für diese Gruppe differenziell valide sind. Daran anschließend wurde untersucht, ob sich itemspezifische Kompetenznachteile durch die sprachlichen Merkmale der Aufgaben erklären lassen. In der vorliegenden Arbeit konnte gezeigt werden, dass die differenzielle Validität der betrachteten Testverfahren für Kinder mit nicht-deutscher Familiensprache insgesamt gering ausgeprägt ist. Ferner wurde festgestellt, dass sich die einzelnen sprachlichen Merkmale der Aufgaben sowohl spezifisch als auch gemeinsam auf die differenzielle Validität auswirken. Der größte Anteil der itemspezifischen Kompetenznachteile wurde durch mehrere Merkmale gemeinsam aufgeklärt. Eine experimentelle Teilstudie zeigte, dass eine sprachliche Vereinfachung nicht geeignet scheint, um die Kompetenznachteile von Kindern mit nicht-deutscher Familiensprache substanziell zu verringern. Ein Vergleich der Effekte sprachlicher Merkmale von Mathematikaufgaben auf die Mathematikleistungen von Kindern mit nicht-deutscher Familiensprache zwischen der dritten und der vierten Klassenstufe ergab, dass sich die sprachliche Komplexität der Aufgaben vor allem für jüngere Grundschulkinder unabhängig von ihrer Familiensprache benachteiligend auswirkte. / Large-scale assessment studies have repeatedly documented performance disadvantages of language minority students in German elementary schools. The substantial achievement gap has led to concerns regarding the validity of large-scale assessment items for language minority students. It may be the case that these performance differences are, in part, due to high language demands of the test items. These items may selectively disadvantage language minority students in the testing situation. This dissertation project investigated the connection between the academic language demands of mathematics test items and the test performance of monolingual students and language minority students. First, it was investigated whether the test items were differentially valid for language minority students. Moreover, the connection between the differential validity and the linguistic complexity of the test items was tested. The findings indicated that overall, differential validity of the examined tests for language minority students was low. However, the test items’ language demands were related to differential validity. The largest proportion of item-specific performance disadvantages was explained by confounded combinations of several linguistic features. Additionally, unique effects of descriptive, lexical, and grammatical features were identified. An experimental study showed that linguistic simplification did not seem to be a promising method to substantially reduce the performance differences between language minority students and German monolingual students. A comparison of differential effects of mathematics items’ language demands for language minority students over two adjacent grade levels indicated that the impact of academic language demands seemed to depend on grade level rather than on language minority student status. Regardless of their home language, younger students seemed to struggle more with linguistically complex test items than older students.