Machine learning systems have become ubiquitous into our society. This has raised concerns about the potential discrimination that these systems might exert due to unconscious bias present in the data, for example regarding gender and race. Whilst this issue has been proposed as an essential subject to be included in the new AI curricula for schools, research has shown that it is a difficult topic to grasp by students. This thesis aims to develop an educational platform tailored to raise the awareness of the societal implications of gender bias in supervised learning. It assesses whether using an explainable model has a positive effect in teaching the impacts of gender bias to preadolescents from 10 to 13 years old. A study was carried out at a school in Stockholm employing an online platform with a classifier incorporating Grad-CAM as the explainability technique that enables it to visually explain its own predictions. The students were divided into two groups differentiated by the use of the explainable model or not. Analysis of the answers demonstrates that preadolescents significantly improve their understanding of the concept of bias in terms of gender discrimination when they interact with the explainable model, highlighting its suitability for educational programs. / Maskininlärningssystemen har blivit allmänt förekommande i vårt samhälle, vilket har lett till oro över den potentiella diskriminering som dessa system kan utöva när det gäller kön och ras. Detta med orsak av det bias som finns i datan. Även om detta problem har föreslagits som ett viktigt ämne som ska ingå i de nya AI-läroplanerna för skolor, har forskning visat att det är ett svårt ämne att förstå för studenter. Detta examensarbete syftar till att utveckla en utbildningsplattform för att öka medvetenhet om de samhälleliga konsekvenserna av könsbias inom övervakad maskinlärning. Det utvärderar huruvida användning av en förklaringsbar modell har en positiv effekt vid inlärning hos ungdomar mellan 10 och 13 år när det kommer till konsekvenserna av könsbias. En studie genomfördes på en skola i Stockholm med hjälp av en onlineplattform som använder en klassificeringsalgoritm med Grad-CAM förklaringsbar teknik som gör det möjligt för den att visuellt förklara sina egna förutsägelser. Eleverna delades in i två grupper som åtskiljdes genom att den ena gruppen använde den förklarbara modellen medan den andra inte gjorde det. Analysen av svaren visar att ungdomar markant förbättrar sin förståelse av könsdiskrimineringsbias när de interagerar med den förklarbara modellen, vilket lyfter fram dess lämplighet för användning inom utbildningsprogram.

[pt] À medida que os modelos de aprendizado de máquina penetram áreas críticas como medicina, sistema de justiça criminal e mercados financeiros, sua opacidade, que impede que as pessoas interpretem a maioria deles, se tornou um problema a ser resolvido. Neste trabalho, apresentamos uma nova taxonomia para classificar qualquer método, abordagem ou estratégia para lidar com o problema da interpretabilidade de modelos de aprendizado de máquina. A taxonomia proposta que preenche uma lacuna existente nas estruturas de taxonomia atuais em relação à percepção subjetiva de diferentes intérpretes sobre um mesmo modelo. Para avaliar a taxonomia proposta, classificamos as contribuições de artigos científicos relevantes da área. / [en] As machine learning models penetrate critical areas like medicine, the criminal justice system, and financial markets, their opacity, which hampers humans ability to interpret most of them, has become a problem to be solved. In this work, we present a new taxonomy to classify any method, approach or strategy to deal with the problem of interpretability of machine learning models. The proposed taxonomy fills a gap in the current taxonomy frameworks regarding the subjective perception of different interpreters about the same model. To evaluate the proposed taxonomy, we have classified the contributions of some relevant scientific articles in the area.

[pt] A perda de percurso (PL) é um parâmetro essencial em modelos de propagação e crucial na determinação da área de cobertura de sistemas móveis. Osmétodos de aprendizado de máquina (ML) tornaram-se ferramentas promissoras para a previsão de propagação de rádio. No entanto, ainda existem algunsdesafios para sua implantação completa, relacionados à seleção das entradasmais significativas do modelo, à compreensão de suas contribuições para asprevisões do modelo e à avaliação adicional da capacidade de generalizaçãopara amostras desconhecidas. Esta tese tem como objetivo projetar modelosde PL baseados em ML otimizados para diferentes aplicações das tecnologias5G e além. Essas aplicações abrangem links de ondas milimétricas (mmWave)para ambientes indoor e outdoor na faixa de frequência de 26,5 a 40 GHz,cobertura de macrocélulas no espectro sub-6 GHz e comunicações veicularesusando campanhas de medições desenvolvidas em CETUC, Rio de Janeiro,Brazil. Vários algoritmos de ML são explorados, como redes neurais artificiais(ANN), regressão de vetor de suporte (SVR), floresta aleatória (RF) e aumentode árvore de gradiente (GTB). Além disso, estendemos dois modelos empíricospara mmWave com previsão de PL melhorada. Propomos uma metodologiapara seleção robusta de modelos de ML e uma metodologia para selecionar ospreditores mais adequados para as máquinas consideradas com base na melhoria de desempenho e na interpretabilidade do modelo. Além disso, para o canalveículo-veículo (V2V), uma técnica de rede neural convolucional (CNN) também é proposta usando uma abordagem de aprendizado por transferência paralidar com conjuntos de dados pequenos. Os testes de generalização propostosmostram a capacidade dos modelos de ML de aprender o padrão entre as entradas do modelo e a PL, mesmo em ambientes e cenários mais desafiadoresde amostras desconhecidas. / [en] Path loss (PL) is an essential parameter in propagation models and critical in determining mobile systems coverage area. Machine learning (ML) methods have become promising tools for radio propagation prediction. However, there are still some challenges for its full deployment, concerning to selection of the most significant model s inputs, understanding their contributions to the model s predictions, and a further evaluation of the generalization capacity for unknown samples. This thesis aims to design optimized ML-based PL models for different applications for the 5G and beyond technologies. These applications encompass millimeter wave (mmWave) links for indoor and outdoor environments in the frequency band from 26.5 to 40 GHz, macrocell coverage in the sub-6 GHz spectrum, and vehicular communications using measurements campaign carried out by the Laboratory of Radio-propagation, CETUC, in Rio de Janeiro, Brazil. Several ML algorithms are exploited, such as artificial neural network (ANN), support vector regression (SVR), random forest (RF), and gradient tree boosting (GTB). Furthermore, we have extended two empirical models for mmWave with improved PL prediction. We proposes a methodology for robust ML model selection and a methodology to select the most suitable predictors for the machines considered based on performance improvement and the model s interpretability. In adittion, for the vehicle-to-vehicle (V2V) channel, a convolutional neural network (CNN) technique is also proposed using a transfer learning approach to deal with small datasets. The generalization tests proposed shows the ability of the ML models to learn the pattern between the model’s inputs and PL, even in more challenging environments and scenarios of unknown samples.

Ce mémoire s’intéresse aux modèles de Markov à variables latentes, une famille de modèles dans laquelle une chaîne de Markov latente régit le comportement d’un processus stochastique observable à travers duquel transparaît une version bruitée de la chaîne cachée. Pouvant être vus comme une généralisation naturelle des modèles de mélange, ces processus stochastiques bivariés ont entre autres démontré leur faculté à capter les dynamiques variables de maintes séries chronologiques et, plus spécifiquement en finance, à reproduire la plupart des faits stylisés des rendements financiers. Nous nous intéressons en particulier aux chaînes de Markov à temps discret et à espace d’états fini, avec l’objectif d’étudier l’apport de leurs reformulations hiérarchiques et de la relaxation de l’hypothèse d’homogénéité de la matrice de transition à la qualité de l’ajustement aux données et des prévisions, ainsi qu’à la reproduction des faits stylisés. Nous présentons à cet effet deux structures hiérarchiques, la première permettant une nouvelle interprétation des relations entre les états de la chaîne, et la seconde permettant de surcroît une plus grande parcimonie dans la paramétrisation de la matrice de transition. Nous nous intéressons de plus à trois extensions non-homogènes, dont deux dépendent de variables observables et une dépend d’une autre variable latente. Nous analysons pour ces modèles la qualité de l’ajustement aux données et des prévisions sur la série des log-rendements du S&P 500 et du taux de change Canada-États-Unis (CADUSD). Nous illustrons de plus la capacité des modèles à reproduire les faits stylisés, et présentons une interprétation des paramètres estimés pour les modèles hiérarchiques et non-homogènes. Les résultats obtenus semblent en général confirmer l’apport potentiel de structures hiérarchiques et des modèles non-homogènes. Ces résultats semblent en particulier suggérer que l’incorporation de dynamiques non-homogènes aux modèles hiérarchiques permette de reproduire plus fidèlement les faits stylisés—même la lente décroissance de l’autocorrélation des rendements centrés en valeur absolue et au carré—et d’améliorer la qualité des prévisions obtenues, tout en conservant la possibilité d’interpréter les paramètres estimés. / This master’s thesis is centered on the Hidden Markov Models, a family of models in which an unobserved Markov chain dictactes the behaviour of an observable stochastic process through which a noisy version of the latent chain is observed. These bivariate stochastic processes that can be seen as a natural generalization of mixture models have shown their ability to capture the varying dynamics of many time series and, more specifically in finance, to reproduce the stylized facts of financial returns. In particular, we are interested in discrete-time Markov chains with finite state spaces, with the objective of studying the contribution of their hierarchical formulations and the relaxation of the homogeneity hypothesis for the transition matrix to the quality of the fit and predictions, as well as the capacity to reproduce the stylized facts. We therefore present two hierarchical structures, the first allowing for new interpretations of the relationships between states of the chain, and the second allowing for a more parsimonious parameterization of the transition matrix. We also present three non-homogeneous models, two of which have transition probabilities dependent on observed explanatory variables, and the third in which the probabilities depend on another latent variable. We first analyze the goodness of fit and the predictive power of our models on the series of log returns of the S&P 500 and the exchange rate between canadian and american currencies (CADUSD). We also illustrate their capacity to reproduce the stylized facts, and present interpretations of the estimated parameters for the hierarchical and non-homogeneous models. In general, our results seem to confirm the contribution of hierarchical and non-homogeneous models to these measures of performance. In particular, these results seem to suggest that the incorporation of non-homogeneous dynamics to a hierarchical structure may allow for a more faithful reproduction of the stylized facts—even the slow decay of the autocorrelation functions of squared and absolute returns—and better predictive power, while still allowing for the interpretation of the estimated parameters.

Insurance is built on the principle that a group of people contributes to a common pool of money which will be used to cover the costs for individuals who suffer from the insured event. In a competitive market, an insurance company will only be profitable if their pricing reflects the covered risks as good as possible. This thesis investigates the recently proposed Combined Actuarial Neural Network (CANN), a model nesting the traditional Generalised Linear Model (GLM) used in insurance pricing into a Neural Network (NN). The main idea of utilising NNs for insurance pricing is to model interactions between features that the GLM is unable to capture. The CANN model is analysed in a commercial insurance setting with respect to two research questions. The first research question, RQ 1, seeks to answer if the CANN model can outperform the underlying GLM with respect to error metrics and actuarial model evaluation tools. The second research question, RQ 2, seeks to identify existing interpretability methods that can be applied to the CANN model and also showcase how they can be applied. The results for RQ 1 show that CANN models are able to consistently outperform the GLM with respect to chosen model evaluation tools. A literature search is conducted to answer RQ 2, identifying interpretability methods that either are applicable or are possibly applicable to the CANN model. One interpretability method is also proposed in this thesis specifically for the CANN model, using model-fitted averages on two-dimensional segments of the data. Three interpretability methods from the literature search and the one proposed in this thesis are demonstrated, illustrating how these may be applied. / Försäkringar bygger på principen att en grupp människor bidrar till en gemensam summa pengar som används för att täcka kostnader för individer som råkar ut för den försäkrade händelsen. I en konkurrensutsatt marknad kommer försäkringsbolag endast vara lönsamma om deras prissättning är så bra som möjligt. Denna uppsats undersöker den nyligen föreslagna Combined Actuarial Neural Network (CANN) modellen som bygger in en Generalised Linear Model (GLM) i ett neuralt nätverk, i en praktiskt och kommersiell försäkringskontext med avseende på två forskningsfrågor. Huvudidén för en CANN modell är att fånga interaktioner mellan variabler, vilket en GLM inte automatiskt kan göra. Forskningsfråga 1 ämnar undersöka huruvida en CANN modell kan prestera bättre än en GLM med avseende på utvalda statistiska prestationsmått och modellutvärderingsverktyg som används av aktuarier. Forskningsfråga 2 ämnar identifiera några tolkningsverktyg som kan appliceras på CANN modellen samt demonstrera hur de kan användas. Resultaten för Forskningsfråga 1 visar att CANN modellen kan prestera bättre än en GLM. En literatursökning genomförs för att svara på Forskningsfråga 2, och ett antal tolkningsverktyg identifieras. Ett tolkningsverktyg föreslås också i denna uppsats specifikt för att tolka CANN modellen. Tre av tolkningsverktygen samt det utvecklade verktyget demonstreras för att visa hur de kan användas för att tolka CANN modellen.

Insurance is built on the principle that a group of people contributes to a common pool of money which will be used to cover the costs for individuals who suffer from the insured event. In a competitive market, an insurance company will only be profitable if their pricing reflects the covered risks as good as possible. This thesis investigates the recently proposed Combined Actuarial Neural Network (CANN), a model nesting the traditional Generalised Linear Model (GLM) used in insurance pricing into a Neural Network (NN). The main idea of utilising NNs for insurance pricing is to model interactions between features that the GLM is unable to capture. The CANN model is analysed in a commercial insurance setting with respect to two research questions. The first research question, RQ 1, seeks to answer if the CANN model can outperform the underlying GLM with respect to error metrics and actuarial model evaluation tools. The second research question, RQ 2, seeks to identify existing interpretability methods that can be applied to the CANN model and also showcase how they can be applied. The results for RQ 1 show that CANN models are able to consistently outperform the GLM with respect to chosen model evaluation tools. A literature search is conducted to answer RQ 2, identifying interpretability methods that either are applicable or are possibly applicable to the CANN model. One interpretability method is also proposed in this thesis specifically for the CANN model, using model-fitted averages on two-dimensional segments of the data. Three interpretability methods from the literature search and the one proposed in this thesis are demonstrated, illustrating how these may be applied. / Försäkringar bygger på principen att en grupp människor bidrar till en gemensam summa pengar som används för att täcka kostnader för individer som råkar ut för den försäkrade händelsen. I en konkurrensutsatt marknad kommer försäkringsbolag endast vara lönsamma om deras prissättning är så bra som möjligt. Denna uppsats undersöker den nyligen föreslagna Combined Actuarial Neural Network (CANN) modellen som bygger in en Generalised Linear Model (GLM) i ett neuralt nätverk, i en praktiskt och kommersiell försäkringskontext med avseende på två forskningsfrågor. Huvudidén för en CANN modell är att fånga interaktioner mellan variabler, vilket en GLM inte automatiskt kan göra. Forskningsfråga 1 ämnar undersöka huruvida en CANN modell kan prestera bättre än en GLM med avseende på utvalda statistiska prestationsmått och modellutvärderingsverktyg som används av aktuarier. Forskningsfråga 2 ämnar identifiera några tolkningsverktyg som kan appliceras på CANN modellen samt demonstrera hur de kan användas. Resultaten för Forskningsfråga 1 visar att CANN modellen kan prestera bättre än en GLM. En literatursökning genomförs för att svara på Forskningsfråga 2, och ett antal tolkningsverktyg identifieras. Ett tolkningsverktyg föreslås också i denna uppsats specifikt för att tolka CANN modellen. Tre av tolkningsverktygen samt det utvecklade verktyget demonstreras för att visa hur de kan användas för att tolka CANN modellen.

This essay critically examines the limitations of artificial intelligence (AI) in achieving human-like understanding and intelligence. Despite significant advancements in AI, such as the development of sophisticated machine learning algorithms and neural networks, current systems fall short in comprehending the cognitive depth and flexibility inherent in human intelligence. Through an exploration of historical and contemporary arguments, including Searle's Chinese Room thought experiment and Dennett's Frame Problem, this essay highlights the inherent differences between human cognition and AI. Central to this analysis is the role of metaphorical thinking and embodied cognition, as articulated by Lakoff and Johnson, which are fundamental to human understanding but absent in AI. Proponents of AGI, like Kurzweil and Bostrom, argue for the potential of AI to surpass human intelligence through recursive self-improvement and technological integration. However, this essay contends that these approaches do not address the core issues of experiential knowledge and contextual awareness. By integrating insights from contemporary scholars like Bender, Koller, Buckner, Thorstad, and Hoffmann, the essay ultimately concludes that AI, while a powerful computational framework, is fundamentally incapaple of replicating the true intelligence and understanding unique to humans.