Background. Code reviewing is a commonly used practice in software development. It refers to the process of reviewing new code changes, commonly before they aremerged with the code base. However, in order to perform the review, developers need to be assigned to that task. The problems with a manual assignment includes a time-consuming selection process; limited pool of known candidates; risk of high reuse of the same reviewers (high workload). Objectives. This thesis aims to attempt to address the above issues with a recommendation system. The idea is to receive feedback from experienced developers in order to expand upon identified reviewer factors; which can be used to determinethe suitability of developers as reviewers for a given change. Also, to develop and implement a solution that uses some of the most promising reviewer factors. The solution can later be deployed and validated through user and reviewer feedback in a real large-scale project. The developed recommendation system is named Carrot. Methods. An improvement case study was conducted at Ericsson. The identification of reviewer factors is found through literature review and semi-structured interviews. Validation of Carrot’s usability was conducted through static analysis,user feedback, and static validation. Results. The results show that Carrot can help identify adequate non-obvious reviewers and be of great assistance to new developers. There are mixed opinions on Carrot’s ability to assist with workload balancing and decrease of review lead time. The recommendations can be performed in a production environment in less than a quarter of a second. Conclusions. The implemented and validated approach indicates possible usefulness in performing recommendations, but could benefit significantly from further improvements. Many of the problems seen with the recommendations seem to be a result of corner-cases that are not handled by the calculations. The problems would benefit considerably from further analysis and testing.

The project aims at exploring the process of designing recommender systems from a users’ perspective. Recommendations are the systems that can help users navigate in the overload of information, that is currently available online. This project focuses on the recommender network of Strossle, which provides article recommendations across various publishers’ websites. User-centered research has been performed to understand the current system and how that influences the users’ perceived experience. The goal was to develop a more engaging content experience for the Strossle recommendation system. This is done by means of participatory design methods. As people tend to use recommendations very sporadic and they often do not really know what they are looking for. The emphasis was on finding the balance between exploratory browsing and navigating towards the users’ preferences. In order to achieve this, a more dynamic widget has been developed that offers navigation in various related topics.

Rekommendationssystem är idag väsentliga för att navigera den enorma mängd produkter tillgängliga via internet. Då social media i form av Twitter vid tidigare tillfällen använts för att generera filmrekommendationer har detta främst varit för att hantera cold-start, ett vanligt drabbande problem för collaborative-filtering. I detta arbete adresseras istället hur top-k rekommendationer påverkas vid integrering av social media data i rekommendationssystemet. För att svara på denna fråga har en prototyp av nytt slag utvecklats inom processmodellen för Design Science. Systemet rankar om top-k rekommendationer baserat på resultatet av social matchning där användares Tweets matchas med nyckelord för filmer genom latent semantic indexing (LSI) similarity. Prototypen evalueras genom experiment som adresserar funktionalitet, noggrannhet, konsekvens och prestanda. Resultatet visar att mätetalen NDCG och MAP för top-k rekommendationer förbättras med social matching jämfört med att enbart använda collaborative filtering. / Recommender systems are a crucial part of navigating the vast number of products on the internet. Social media, in the form of Twitter microblogs, has been previously used to produce movie recommendations, yet this has mainly been to solve cold-start, a common problem in collaborative filtering environments. This work addresses how top-k recommendations in a collaborative filtering environment are affected when augmented with social media data. To answer this question a novel prototype is developed following a design science process model. This system re-ranks top-k recommendations based on a social matching process where Tweets are matched with movie keywords through latent semantic indexing (LSI) similarity. The prototype is evaluated through experiments regarding functionality, accuracy, consistency, and performance. The results show that NDCG and MAP metrics of the top-k recommendations improve with social matching compared to only using the collaborative filtering algorithms.

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第21390号 / 情博第676号 / 新制||情||117(附属図書館) / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 鹿島 久嗣, 教授 山本 章博, 教授 下平 英寿 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Information Technology (IT) has evolved over the decades, where its role and impact have transitioned from being a tactical tool to a more strategic one for driving business strategies to transform organizations. The right alignment between IT strategy and business has become a compelling factor for Chief Information Officer (CIO)s where misalignment could lead to a degradation of organization performance and slow them down in the landscape of market competition. Enterprise Architecture (EA) in practice is one of the approaches where this alignment can be achieved. Through Enterprise Modeling (EM), EA supports all the stakeholders (Business and IT) of an organization to have a common understanding for communication of information, making decisions, and coordinating actions. EM results in EA models that are composed of enterprise components and relationships, that are stored in a repository. Over time, the repository grows as architects design new models or update existing models with complex structures. This opens up new avenues of research where the data can be used to provide intelligence. One such field is Recommender Systems (RS) which are software tools and techniques to provide meaningful suggestions to the user. RS can take different forms in the EM domain. For instance, identifying semantically similar components [8] or modeling assistance during the EM process [40]. Each form of recommendation can be enhanced with sophisticated recommendation models over time implemented in different technologies. We focus on the latter problem in this thesis where we implement a recommender framework that contains a robust architecture that easily integrates with different RS. We present the robustness and the ease of integration criteria for the framework. The framework is distributed as a plugin for Archi, a modeling toolkit used by Enterprise Architects to create EA models. / Information Technology (IT) har utvecklats under decennierna, där dess roll och inverkan har övergått från att vara ett taktiskt verktyg till ett mer strategiskt verktyg för att driva affärsstrategier för att transformera organisationer. Den rätta anpassningen mellan IT-strategin och verksamheten har blivit en övertygande faktor för Chief Information Officer (CIO):er där felanpassning kan leda till en försämring av organisationens prestanda och sakta ner dem i konkurrensen på marknaden. Enterprise Architecture (EA) är i praktiken ett av tillvägagångssätten där denna anpassning kan uppnås. Genom Enterprise Modeling (EM) stöder EA alla intressenter (Business och IT) i en organisation att ha en gemensam förståelse för kommunikation av information, fatta beslut och samordna åtgärder. EM resulterar i EA-modeller som är sammansatta av företagskomponenter och relationer, som lagras i ett arkiv. Med tiden växer förvaret i takt med att arkitekter designar nya modeller eller uppdaterar befintliga modeller med komplexa strukturer. Detta öppnar för nya forskningsvägar där data kan användas för att tillhandahålla intelligens. Ett sådant fält är Recommender Systems (RS) som är mjukvaruverktyg och tekniker för att ge meningsfulla förslag till användaren. RS kan ha olika former i EM-domänen. Till exempel identifiera semantiskt liknande komponenter [8] eller modelleringshjälp under EM-processen [40]. Varje form av rekommendation kan förbättras med sofistikerade rekommendationsmodeller över tid implementerade i olika teknologier. Vi fokuserar på det senare problemet i den här avhandlingen där vi implementerar ett recommender-ramverk som innehåller en robust arkitektur som enkelt integreras med olika RS. Vi presenterar robustheten och enkelheten i integrationskriterierna för ramverket. Ramverket distribueras som en plugin för Archi, en verktygssats för modellering som används av Enterprise Architects för att skapa EA-modeller.

Popular social media platforms use recommender systems (RS) to improve users’ experience. However, these systems pose potential issues including political polarization and fake news spread. In addition, social media RS may raise many ethical issues, such as Privacy, Opacity, Fairness and Social Effects, regarding the collection and processing of data. The research problem of this study is the lack of knowledge regarding user concerns toward these issues. Therefore, this study attempts to address the research gap by studying user attitudes toward the aforementioned ethical issues. The main research is: What are users’ concerns about ethical issues of using recommender systems in social media platforms? For this study, a survey was chosen as the research strategy. A questionnaire was developed and distributed online to collect data. The data analysis is based on descriptive statistics using measures of central tendency by finding the median (the middle point). Key findings indicate that users are familiar and concerned with the major issues. In particular, users are concerned about how and what types of private data are collected. However, users are most concerned about the spread of fake news. Users also express concern about political polarization, although they do not believe they are personally affected. However, users would prefer that recommender algorithms use less private data, even if it compromises news feed quality. The study has limitations including a geographic focus on Europe and demographics. Nevertheless, the findings offer insights into user concerns about RS, paving the way for future research.

In this report we evaluated how the complexity of a Convolutional Neural Network (CNN), in terms of number of filters, size of filters and dropout, affects the performance on the rating prediction accuracy in a tightly coupled hybrid recommender system. We also evaluated the effect on the rating prediction accuracy for pretrained CNNs in comparison to non-pretrained CNNs. We found that a less complex model, i.e. smaller filters and less number of filters, showed trends of better performance. Less regularization, in terms of dropout, had trends of better performance for the less complex models. Regarding the comparison of the pretrained models and non-pretrained models the experimental results were almost identical for the two denser datasets while pretraining had slightly worse performance on the sparsest dataset. / I denna rapport utvärderade vi komplexiteten på ett neuralt faltningsnätverk (eng. Convolutional Neural Network) i form av antal filter, storleken på filtren och regularisering, i form av avhopp (eng. dropout), för att se hur dessa hyperparametrar påverkade träffsäkerheten för rekommendationer i ett hybridrekommendationssystem. Vi utvärderade även hur förträning av det neurala faltningsnätverket påverkade träffsäkerheten för rekommendationer i jämförelse med ett icke förtränat neuralt faltningsnätverk. Resultaten visade trender på att en mindre komplex modell, det vill säga mindre och färre filter, gav bättre resultat. Även mindre regularisering, i form av avhopp, gav bättre resultat för mindre komplexa modeller. Gällande jämförelsen med förtränade modeller och icke förtränade modeller visade de experimentella resultaten nästan ingen skillnad för de två kompaktare dataseten medan förträning gav lite sämre resultat på det glesaste datasetet.

Recommender systems are information filtering tools that discover potential matchings between users and items and benefit both parties. This benefit can be considered a social resource that should be equitably allocated across users and items, especially in critical domains such as education and employment. Biases and unfairness in recommendations raise both ethical and legal concerns. In this dissertation, we investigate the concept of unfairness in the context of recommender systems. In particular, we study appropriate unfairness evaluation metrics, examine the relation between bias in recommender models and inequality in the underlying population, as well as propose effective unfairness mitigation approaches. We start with exploring the implication of fairness in recommendation and formulating unfairness evaluation metrics. We focus on the task of rating prediction. We identify the insufficiency of demographic parity for scenarios where the target variable is justifiably dependent on demographic features. Then we propose an alternative set of unfairness metrics that measured based on how much the average predicted ratings deviate from average true ratings. We also reduce these unfairness in matrix factorization (MF) models by explicitly adding them as penalty terms to learning objectives. Next, we target a form of unfairness in matrix factorization models observed as disparate model performance across user groups. We identify four types of biases in the training data that contribute to higher subpopulation error. Then we propose personalized regularization learning (PRL), which learns personalized regularization parameters that directly address the data biases. PRL poses the hyperparameter search problem as a secondary learning task. It enables back-propagation to learn the personalized regularization parameters by leveraging the closed-form solutions of alternating least squares (ALS) to solve MF. Furthermore, the learned parameters are interpretable and provide insights into how fairness is improved. Third, we conduct theoretical analysis on the long-term dynamics of inequality in the underlying population, in terms of the fitting between users and items. We view the task of recommendation as solving a set of classification problems through threshold policies. We mathematically formulate the transition dynamics of user-item fit in one step of recommendation. Then we prove that a system with the formulated dynamics always has at least one equilibrium, and we provide sufficient conditions for the equilibrium to be unique. We also show that, depending on the item category relationships and the recommendation policies, recommendations in one item category can reshape the user-item fit in another item category. To summarize, in this research, we examine different fairness criteria in rating prediction and recommendation, study the dynamic of interactions between recommender systems and users, and propose mitigation methods to promote fairness and equality. / Doctor of Philosophy / Recommender systems are information filtering tools that discover potential matching between users and items. However, a recommender system, if not properly built, may not treat users and items equitably, which raises ethical and legal concerns. In this research, we explore the implication of fairness in the context of recommender systems, study the relation between unfairness in recommender output and inequality in the underlying population, and propose effective unfairness mitigation approaches. We start with finding unfairness metrics appropriate for recommender systems. We focus on the task of rating prediction, which is a crucial step in recommender systems. We propose a set of unfairness metrics measured as the disparity in how much predictions deviate from the ground truth ratings. We also offer a mitigation method to reduce these forms of unfairness in matrix factorization models Next, we look deeper into the factors that contribute to error-based unfairness in matrix factorization models and identify four types of biases that contribute to higher subpopulation error. Then we propose personalized regularization learning (PRL), which is a mitigation strategy that learns personalized regularization parameters to directly addresses data biases. The learned per-user regularization parameters are interpretable and provide insight into how fairness is improved. Third, we conduct a theoretical study on the long-term dynamics of the inequality in the fitting (e.g., interest, qualification, etc.) between users and items. We first mathematically formulate the transition dynamics of user-item fit in one step of recommendation. Then we discuss the existence and uniqueness of system equilibrium as the one-step dynamics repeat. We also show that depending on the relation between item categories and the recommendation policies (unconstrained or fair), recommendations in one item category can reshape the user-item fit in another item category. In summary, we examine different fairness criteria in rating prediction and recommendation, study the dynamics of interactions between recommender systems and users, and propose mitigation methods to promote fairness and equality.

Indiana University-Purdue University Indianapolis (IUPUI) / The dramatic improvement in information and communication technology (ICT) has made an evolution in learning management systems (LMS). The rapid growth in LMSs has caused users to demand more advanced, automated, and intelligent services. CourseNet working is a next-generation LMS adopting machine learning to add personalization, gamifi cation, and more dynamics to the system. This work tries to come up with two recommender systems that can help improve CourseNetworking services. The first one is a social recommender system helping CourseNetworking to track user interests and give more relevant recommendations. Recently, graph neural network (GNN) techniques have been employed in social recommender systems due to their high success in graph representation learning, including social network graphs. Despite the rapid advances in recommender systems performance, dealing with the dynamic property of the social network data is one of the key challenges that is remained to be addressed. In this research, a novel method is presented that provides social recommendations by incorporating the dynamic property of social network data in a heterogeneous graph by supplementing the graph with time span nodes that are used to define users long-term and short-term preferences over time. The second service that is proposed to add to Rumi services is a hashtag recommendation system that can help users label their posts quickly resulting in improved searchability of content. In recent years, several hashtag recommendation methods are proposed and de veloped to speed up processing of the texts and quickly find out the critical phrases. The methods use different approaches and techniques to obtain critical information from a large amount of data. This work investigates the efficiency of unsupervised keyword extraction methods for hashtag recommendation and recommends the one with the best performance to use in a hashtag recommender system.

This thesis explores potential strategies to enhance user control and satisfaction within the movie selection process, with a particular focus on the utilization of conversational generative artificial intelligence, such as ChatGPT, for personalized movie recommendations. The study adopts a qualitative user-centered design thinking approach, aiming to compre-hensively understand user needs, goals, and behavior. In-depth interviews were conducted, utilizing the "Thinking aloud"method and trigger materials to elicit rich user feedback. Participants interacted with ChatGPT and various prototypes, providing valuable insights into their experiences. The study found that participants felt more in control when given the opportunity to specify wishes. In addition, the users found that the experience of receiving recommendations through ChatGPT was more satisfying than their usual way of receiving recommendations for movies. Furthermore, participants expressed a desire for additional information about recommended movies and more novel suggestions. The prototypes, designed as triggers for user feedback, were generally well-received, providing an engaging and fun user experience. Despite some participants expressing challenges in specifying movie choices based on an emotion, this new approach to movie selection was viewed positively. Despite limitations concerning the study’s validity, reliability, and testing situation, the findings suggest the potential of generative artificial intelligence conversational agents in enhancing the movie selection process. It is concluded that iterative design improvements and further research is necessary to fully leverage the potential of natural language processing technologies in recommendation systems. The study serves as a preliminary investigation into improving movie recommendations using generative artificial intelligence and offers valuable insights for future developments.