Publikationer från Stockholms universitet
Not a member yet
31237 research outputs found
Sort by
Friendship as a Pathway: The Role of Peer Employment in Migrants’ Labor Market Integration : Social Networks and Employment Across Migration, and Generational Status
No full textThis thesis investigates how friends’ employment status influences individual employment outcomes, comparing migrant populations in the United Kingdom with natives and first-generation migrants with their children. Drawing on Waves 3 and 6 of the Understanding Society survey and employing survey-weighted logistic regression models and their corresponding Average Marginal Effects, the study examines whether and how peer employment effects vary across different social and migration groups. The analysis proceeds in four stepwise models, followed by three interaction models that explore differential effects between migrants and natives, between first- and second-generation migrants, and across varying durations of residence among first-generation migrants. The findings reveal that friends’ employment exerts a strong and enduring influence on individuals’ likelihood of being employed, supporting theoretical frameworks of social contagion and network capital. For migrants, particularly recent and first-generation arrivals, these peer effects are amplified, serving as an important bridge into the host-country labor market. Over time and across generations, as migrants integrate and diversify their social networks, their reliance on friends’ employment diminishes but remains important
Artificial Intelligence in Project Selection in Process Industries: Enhancing Prioritization, Risk Management, and Decision-Making
No full textArtificial intelligence (AI) is increasingly applied in project selection within process industries such as oil and gas, mining, and chemical manufacturing, yet its specific impact on prioritization, risk management, and decision-making remains underexplored. This thesis addresses this gap through a qualitative study that is a systematic literature review with manual content analysis of 36 peer-reviewed articles. The findings reveal that AI enhances project selection by enabling multi-criteria evaluation, improving risk identification and mitigation, and supporting faster, data-driven decisions. Importantly, the study distinguishes between AI integration (technical embedding in organizational systems) and AI utilization (practical application and adoption), emphasizing the need for human–AI collaboration to achieve meaningful benefits. The research contributes theoretically by framing AI as both a technical enabler and a strategic decision-support tool, and practically by offering managers insights on leveraging AI to align project portfolios with corporate strategy while improving operational resilience. Future research is encouraged to validate these findings through empirical case studies and to examine in greater depth the ethical dimensions of AI-driven decision-making
The Effects of Fluoride Exposure on Child Development : Evidence from a Water Fluoridation Experiment
No full textAlthough water fluoridation is widely implemented to promote dental health, causal evidence of its broader developmental effects remains limited. This study exploits a water fluoridation experiment in Norrköping, Sweden (1952–1962), to investigate the long-term effects of childhood fluoride exposure on human capital outcomes. The results show that early-life exposure to fluoridated water is associated with reduced non-cognitive ability at age 18 and a lower likelihood of high school graduation. These findings highlight the need for careful reassessment of what constitutes safe fluoride levels in public water supplies
DFET modeling: A Case of Industrial Control Systems Threat Modeling - Improving Industrial Control System Security Postures by Combining Digital Forensics and Threat Modeling
No full textIndustrial control systems (ICS) are a term used to describe control and automation systems in critical infrastructure such as power generation and distribution, transportation, water treatment, and utilities. These control systems and devices were not built with security in mind, as they were often separated, “air-gapped”, from an organization’s enterprise network. As the internet made its entry, the air gap started diminishing. In recent years, due to technological development, these systems have become the targets of malicious actors who want to attack these systems for purposes such as espionage and sabotage. This calls for the incorporation of digital forensic readiness in such systems. There is currently a gap in the forensic readiness as these systems lack built-in capabilities to facilitate forensic investigations before and during attacks. The use of threat modeling can greatly help organizations to improve their security posture and decrease the probability of a cyberattack. The research questions that this papers seeked to answer were “What research has been conducted to combine digital forensics and digital forensic readiness with threat modeling?” and “How can digital forensic principles be integrated into threat modeling to improve digital forensic readiness in ICS?” To answer the first question, an extensive systematic literature study was conducted starting with 3,114 unique articles that were systematically reduced to 68 articles for a final analysis. They were analyzed to see whether forensic readiness was utilized in the presented solutions, and whether the forensic capabilities had been integrated into their model (i.e. threat model). The articles covered different industries and areas including ICS, cloud computing, medical technology, and transportation. The results showed that only 36 of the articles had in some way incorporated digital forensic (readiness) in their threat models. Two articles by Asif Iqbal discussed the development and application of the Digital Forensic Evidence-based Threat (DFET) model, proved to be the most comprehensive model, as it shows that digital evidence is created along each interaction that an attacker performs on a system under attack. To answer the second question, the author showed how the DFET model can be applied to the different stages of the digital forensic investigation process, using a hypothetical scenario where an attacker would compromise, and subsequently delete evidence from an FTP server that allows for login without credentials. The model showed that by applying the DFET model, it would be possible to anticipate what forensic artifacts could be created, as well as what potential anti-forensic measures an attacker may take. The DFET model has a huge potential to help organizations and ICSs to improve their security posture, by taking measures to ensure that pre- and post-attacks are taken into account
Terminal Circulation : A Critical Theory of the Algorimage
No full textThis thesis constructs and develops the concept of the algorimage as a historically-specific image form that is dominant in contemporary society. It critically examines the appearance of algorithmically circulated images, both in terms of their socio-technical conditions of appearance, as well as the appearance of these conditions as instantiated in concrete audiovisual images and abstract mental images of “the algorithm.” The concept aims to address a particular problematic: the increasing visibility of algorimages and the increasing invisibility of their microtemporal, algorithmic operations within the shift of capital’s center of gravity toward the sphere of circulation—logistics, finance, social media, and unproductive labor. The algorimage is theorized as a form of appearance, or, a social relation between a growing surplus of images circulating across a surfeit of screens in derivative, recursive, and compressed content formats on the one hand, and the declining surplus value of a stagnating capitalism on the other. This thesis mobilizes insights from an interdisciplinary set of resources: image and film theory, turns in cinema and media studies toward materialist analyses of logistical media and infrastructures, media archaeological digs into the technical substrates of media and techniques of mediation, and the Marxist critique of political economy. Chapter 1 introduces the theoretical and methodological framework deployed for the theorization of the algorimage. It first accounts for what an image is and does; then synthesizes various materialist approaches to the technical and the social, and their reciprocal mediation; before concluding with a critique of prior theorizations of the relation between digital images and contemporary capital in terms of an “iconomy.” Chapter 2 is an analysis of Twitter/X’s technical infrastructure as a data supply chain across which the algorimage circulates as a container technology for the valorization of data to be realized as profit. The continual update of content moves in accordance with logistical imperatives of speed-up and efficiency, which in turn formats, and is formatted by, algorimages, as shown in the analysis of so-called “object-labeling”memes. Chapter 3 analyzes the short-form and the stream as two paradigmatic algorimage formats. Together they comprise a continuous surplus of clips which instantiate a dialectical duration that oscillates between the respective micro and macro temporalities of algorithmic and capital circulation. Chapter 4 is a case study of the “reaction video” genre as an imaging of recursive recommendation algorithms. By performing reaction, these videos also perform their algorithms with the aim of validating an apparent correlation between effective computation and the “affective labor” of producing content for a platform economy that runs on the reactions of its users. In chapter 5, the abstraction of the algorimage is concretized as part of the fabric of social movements through images taken from, and of, the George Floyd Uprisings of 2020. It aims to theorize how the conditions from which the uprisings emerged are the same as conditions of appearance of the algorimage—the historical development of the antagonism between labor and capital, and the change in the primary form and subject of struggle from the strike and the worker to the riot and the surplus population
NLP-based Deepfake Text Detection - Identifying AI-generated Fraudulent Text
No full textIntroduction: The increasing accessibility of large language models (LLMs) such as GPT-4 has raised significant concerns about their misuse in generating fraudulent content. In particular, fraudulent information generated by artificial intelligence (AI) has become more complex, more convincing, and difficult to detect through traditional rule-based or keyword-driven filtering systems. Because fraudulent texts often rely on well-structured, human-like languages, they present new challenges to content auditing and security systems. Research Question: This study investigates the following question: How effective are current AI text detection tools in identifying LLM generated fraudulent texts? It aims to compare the performance of different detection tools to determine which performs best in identifying AI-generated fraudulent content. Method: To address this question, an experimental approach was adopted. A balanced dataset of AI-generated and human-written fraudulent texts was constructed. The AI texts were produced using ChatGPT, and the human-written samples were sourced from publicly available datasets. Three representative detection tools were selected for evaluation: Giant Language Model Test Room (GLTR), ZeroGPT and DetectGPT. Each tool was applied to the dataset, and its performance was evaluated using standard classification metrics including accuracy, precision, recall and F1-score. Results: The results demonstrate that GLTR provided the most balanced detection of AI-generated fraudulent texts, with accuracy, precision, and recall all close to 0.78. DetectGPT also performed strongly, maintaining recall above 0.70 and accuracy of 0.71. In contrast, ZeroGPT, while achieving perfect precision (1.00), detected only a small fraction of AI-generated texts, with recall of 0.18. Performance further varied by text length: medium-length texts yielded the highest detection rates, while short texts produced the most misclassifications. Combined, these findings highlight the different strengths and weaknesses of current AI text detection tools and underscore the limitations of relying on a single method for identifying fraudulent content. Discussion:These findings suggest that while current AI text detection tools exhibit promising performance under controlled conditions, their effectiveness in detecting LLM-generated fraudulent texts differs greatly across contexts. GLTR and DetectGPT show more balanced detection, but their accuracy drops with shorter texts or when the content is adversarially altered. ZeroGPT, while reaching perfect precision, misses most fraudulent AI-generated texts. These results highlight the need for multi-strategy detection systems and raise concerns about the reliability of existing detectors in high-risk environments such as phishing or impersonation scams
Victorian oppression and vampiric liberation
No full textSedan Bram Stoker år 1897 publicerade skräckromanen Dracula har boken haft ett stort kulturellt inflytande och nytolkats i otaliga filmatiseringar. Berättelsens huvudsakliga kvinnoroller är Mina, Lucy och tre vampyrsystrar som lever i Draculas slott. Diskussionen kring dessa karaktärer, främst Mina och Lucy, har länge varit splittrad. De kopplas ofta till den nya kvinnan, som uppkom i den viktorianska feministrörelse där kvinnor i allt högre grad sökte sig till mansdominerade sfärer. Med utgångspunkt i tidigare analyser av dessa karaktärer syftar uppsatsen till att undersöka hur Mina, Lucy och vampyrsystrarna porträtteras i ett urval av de filmatiseringar som inspirerats av Stokers roman. Detta sker med bas i kategoriseringarna kvinnan som offer och kvinnan som vampyr. Verken som studeras här är följande; Nosferatu, eine Symphonie des Grauens (F. W. Murnau, 1922), Dracula (Tod Browning, 1931), Dracula (Terence Fisher, 1958), Nosferatu - Phantom der Nacht (Werner Herzog, 1979), Dracula (John Badham, 1979), Bram Stoker’s Dracula (Francis Ford Coppola, 1992) och Nosferatu (Robert Eggers, 2024). Det teoretiska ramverk som appliceras på dessa filmer utgörs av Barbara Creeds The Monstrous-Feminine och Laura Mulveys “Visual Pleasure and Narrative Cinema”. Dessa används för att diskutera gestaltningen av kvinnliga offer och kvinnliga vampyrer samt frågor som rör passivitet, handlingskraft och den estetiska stilens upprätthållande av könsroller. Studiens resultat visar att Dracula symboliserar kvinnors förtryckta begär, vilket är centralt för tolkningen av adaptionerna. Dessutom är blodsugandet en erotisk akt, där vampyren med sina falliska huggtänder penetrerar kvinnans hud och samtidigt frigör henne sexuellt. Fallossymboler används genomgående av männen i filmerna för att markera kvinnan som deras egendom. Draculas adaptioner etablerar i princip alltid den manliga karaktären som protagonist och åskådaren projicerar då sitt idealego på honom. Därmed tillåts publiken också etablera ett ägandeskap över kvinnan som dessutom stiliseras efter deras blick. De mänskliga männen spelar rollen av auktoritära figurer vars huvudsyfte är att upprätthålla den patriarkala ordningen. När kvinnan transformeras till en vampyr går hon från att vara ett passivt offer till att bli sexuellt aggressiv. Hon framställs också som abjekt och i totalt uppror mot viktorianska könsroller. För att återta kontrollen över vampyrkvinnan och återställa hennes passivitet använder den auktoritära mannen falliskt penetrerande objekt såsom träpålen.
The meaning of seniority for learning in a hybrid workplace : A qualitative study on the impact of hybrid work on learning during the induction process of new employees of different seniority at an insurance company
No full textStudien är en kvalitativ undersökning vars syfte är att beskriva och belysa hur en hybrid introduktion för nyanställda i försäkringsbranschen påverkar villkoren för lärande, samt huruvida senioritet i yrket kan vara en påverkansfaktor. Studien bygger på åtta semistrukturerade intervjuer med nyanställda på ett försäkringsbolag, där kraven var en anställning på minst en månad och max ett år. Resultatet presenteras utifrån teorin om situerat lärande och visar att den hybrida introduktionen påverkar de juniora mer än de seniora, det här eftersom de juniora anställdes utan tidigare kunskap och var därför perifera deltagare i både sakkunskap och gemenskap. Seniora medarbetare är mer centrala deltagare vid nyanställning eftersom de har mer sakkunskap, vilket ger dem en mer upplevd central roll som främjar deras lärande i jämförelse med de juniora som är mer beroende av de sociala aspekterna för utveckling. Sammanfattningsvis tyder empirin att det finns ett värde för både juniora och seniora att ses fysiskt på kontoret för att skapa goda förutsättningar för det individuella lärandet och organisationsutvecklingen. The study is a qualitative investigation whose purpose is to describe and highlight how a hybrid introduction for new employees in the insurance industry affects the conditions for learning, as well as whether seniority in the profession can be an impacting factor. The study is based on eight semi- structured interviews with new employees at an international insurance company, where the requirements were an employment of at least one month and maximum of one year. The data has been analyzed based on the theory of situated learning in a community of practice and the result shows that the hybrid introduction affects the junior more than the senior, that is because the junior employees were hired without any previous knowledge and were therefore more peripheral participants in the community of practice in both expertise and community. Senior new employees are more central participants because they have more expertise, which promotes their learning compared to the juniors who are more dependent on the social aspects for development. In summary, the empirical evidence shows that there is value for both juniors and seniors to see each other physically in the office to create good conditions for individual learning but also organizational development.
Reliability and Trust: A Qualitative Study Exploring Perceptions of Gen. AI
No full textThe use of generative artificial intelligence, such as ChatGPT, has vastly increased in recent years. Through its capability to aid you in nearly any task that can be formulated into a sentence, many people have come to rely on these types of tools in their everyday life, even without fully understanding its underlying mechanisms. This thesis explores how users perceive and interpret ChatGPT outputs, and how these interpretations shape their trust and verification behaviors in this tool. Through six semi structured interviews with ChatGPT users, a thematic analysis identified a plethora of variables affecting trust and verification behaviors. Through analyzing these variables this study found that trust is highly dynamic and context dependent, often being influenced by surface level cues such as output presentation and language rather than factual accuracy. Domain knowledge played a significant role in determining verification ability. These findings highlight user vulnerabilities to misinformation and emphasize the need for improved transparency and user education in AI systems
