Högskolan i Skövde

ContextCybersecurity incidents targeting personal information are becoming more open with organizations remaining largely reliant on digital systems, cloud services, and data-driven operations. The influence of previous studies on such domains as security responses to security incidents, system recovery, and cybersecurity governance is relatively few, but less emphasis is on how organizations integrate the privacy and secure handling of personal data in post-incident review when a cyber incident of this type involving personal data has taken place.

ObjectiveThis study investigates the extent to which organizations use privacy and personal data protection in post-incident reviews following cybersecurity incidents involving personal data. The study` focuses on organizational assessments of privacy threats, governance organization, organizational learning initiatives, and privacy challenges in both incident response and reviewing.

MethodThe study used a qualitative method, including semi-structured interviews conducted with employees from both cybersecurity incident response and post-incident review of two organizations. Data were thematically analyzed using Braun and Clarke’s (2006) method. It described scrutiny of organization’s privacy risk assessment model, governance design, technical-non-technical team collaboration, incident reporting mechanisms, and organizational learning.

ConclusionIt revealed the fact that the protection of privacy and personal data among information workers in organizations is widely accepted as the basis of a safe cybersecurity incident. But the matter of privacy is generally seen as secondary to technical containment and operational recovery. For the analysis, some challenges were identified such as low depth of team work, an unstructured approach towards assessing privacy risks, the presence of technical bias in review of their post-incident process, as well as informal learning from an organizational viewpoint. The result showed that organizations would benefit from taking a more interdisciplinary approach that weaves together aspects of technical, organizational, governance and privacy dimensions of incident response and post-incident review processes.

As organisations move more of their operations online, they face a growing and dangerous threat double extortion ransomware. This is not just about hackers locking files; they are now stealing sensitive data and threatening to leak it publicly to force a payment. While many companies focus on preventing these attacks with technical tools, this research study shows that prevention alone is not enough. Instead, organisations must become cyber resilient capable of withstanding and recovering from an attack while keeping their business running.

This research study investigated how Swedish organisations strengthen their cyber resilience and the real-world hurdles they face in the process. By reviewing 21 academic papers (S1–S21) and interviewing four cybersecurity experts (P1–P4), a clear pattern emerged. There is a significant gap between security theory and daily practice.

The findings reveal that while advanced technical tools like AI-driven detection exist, they are often let down by human factors. Participants described a coordination nightmare where technical teams and management struggle to communicate during a crisis, leading to role confusion and delayed decisions. Furthermore, smaller teams often lack the personnel and budget to fight off highly professionalized cybercriminal groups.

In conclusion, this research study highlights that true cyber resilience requires more than just better software. It demands a shift in organisational culture and much better coordination between the people, processes, and technology involved in the fight against double extortion ransomware attacks.

Context: For the protection of information assets and following laws and agreements, information security risk management is very important for the organization. Considering the existing ways of dealing with it are mainly dependent on human analysis and professionals experience. Security-related data that is mostly unstructured data, such as reports on incidents, audit results, policy papers, and system logs, is created because of digital infrastructures, and using traditional risk assessment methods makes it very difficult to meet the requirements of quality and complexity. It has now become very easy to get structured information from unstructured text because of the new developments in AI, especially in natural language processing and large language models (LLMs).

Still, there are additional issues that need to be thought about when using AI based LLMs in information security risk management. These include bad data, model bias, being able to explain things, integrating with current governance, risk, and compliance (GRC) frameworks, and keeping private corporate data safe.

Method: This report is a systematic literature review based on PRISMA 2020 guideline of peer-reviewed studies that look into how AI based methods mainly focusing on LLM-based methods are used to turn unstructured data about security into structured data that can be used for assessing information security risk, as well as how problems with integrating these methods with GRC processes are dealt with.

In the report, academic libraries such as Scopus, IEEE Xplore, ScienceDirect, and AMC DL to find, screen, and put together relevant work on how AI can be used to find risks, analyze and rate them, choose controls, and keep checks on them. Many studies indicate that AI based methods specifically LLMs approaches can help with different parts of the risk management lifecycle, such as finding weaknesses and threats, ranking the incidents on the basis of severity, linking data to the risk registers, and making comments about risk, but there is still limitations when trying to integrate into frameworks like ISO/IEC 27005 and NIST SP 800-30.

Moreover, for enhancing the explainability, decreasing the bias, and increasing the privacy of AI outputs, there are researches which provide different methods. But for maintaining accountability and transparency, it lacks developed methods to embed these into formal Governance, Risk, and Compliance (GRC) workflows.

This report identifies major research gaps and proposes future research areas for the utilization of AI-based solutions in corporate information security risk management, assuring safety, rationality, and legal compliance.

The findings explore the employee risk perception in cloud-based Shadow IT use and implications of this to GDPR in small organizations. The study was conducted using a mixed-methods approach consisting of a Systematic Literature Review (SLR) and semi-structured interviews. The findings suggest that Shadow IT is a widespread phenomenon in small organizations where the main drivers of use are the convenience, efficiency and ease of use of the unofficial systems compared to the advantages of using officially approved systems. Unauthorized cloud services are sometimes taken up because of functionality limitations of existing tools and time constraints. Simultaneously, the risks related to these practices that are associated with GDPR are often underestimated. Some people notice that certain individuals are aware of the compliance risks, but some don’t believe that these risks will occur, particularly in smaller companies. The results also indicate that the responses to Shadow IT have been extremely informal and are not regulated by any systematic measures. The importance of the risk perception is in the development of the behavior: the higher the level of awareness, the more careful practices will be, the lower the level of awareness, the more powerful the effect of normalizing the non-conforming behavior. In the context of both behavioral and regulatory aspects, this paper details the importance of usability and risk awareness, both of which are vital to improving GDPR compliance. The results will help to understand how to achieve the balance between productivity and regulatory requirements in the scenario of cloud-based technologies.

Cybersecurity awareness training increasingly relies on analog games such as card games and tabletop exercises to engage non-expert audiences in interactive learning. Despite the growing use of these formats, no validated short-form evaluation instrument exists that is specifically designed to measure their effectiveness in practical training settings. Existing instruments are either too long for deployment in short sessions, developed exclusively for digital game contexts, or used without formal validation. This project addressed that gap by developing and pilot-testing a short-form evaluation instrument for analog cybersecurity awareness games. 

The project followed a four-phase sequential design. A structured scoping review of 21 peer-reviewed papers confirmed the research gap and identified six constructs for inclusion: perceived knowledge gain, self-efficacy, behavioural intention, enjoyability, engagement, and perceived usability. An evidence-based instrument design phase produced a seven-item Version 1.0 questionnaire, with each item adapted from validated measures identified in the literature. Face validity was tested through cognitive interviews with four participants using the think-aloud method, resulting in one substantive revision to produce Version 2.0. A feasibility pilot study was then conducted across two sessions in Sweden and India, involving 60 participants who played the Elevation of Privilege card game before completing either the short-form instrument or a 16-item comparison version in a between-subjects design. 

The short-form instrument met all four pre-specified feasibility thresholds: a median completion time of 41 seconds (SD = 9 seconds), a completion rate of 100%, an item nonresponse rate of 0.9%, and a perceived burden rating of Just right from 90% of participants. An exploratory internal consistency analysis yielded a Cronbach’s alpha of 0.74, meeting the commonly used threshold for acceptable reliability in earlystage instrument development. These findings provide initial evidence that a shortform questionnaire of seven items can achieve practical usability while demonstrating acceptable face validity and internal consistency in the context of analog cybersecurity awareness game evaluation. 

Studiens syfte är att undersöka efterlevnaden av formella krav för en central och kritisk roll inom informationssäkerhet och ta fram faktorer att beakta vid utformning av utbildningsprogram. Studien fokuserar på verklig efterlevnad i praktiken av de formella krav och styrdokument som omfattar rollen. En fallstudie med tre datainsamlingstekniker i form av analys av styrdokument, enkätundersökning och semistrukturerade intervjuer resulterade i insamlat material från sex systemägare.

Resultatet visar på en låg efterlevnadsgrad av flera formella krav. Majoriteten av systemägarna delar liknande uppfattningar kring orsakerna. Resultatet presenterar flera faktorer som identifierats som orsaker till bristande efterlevnad. Dessa faktorer innefattar otydligt ansvar, bristande tillämpbarhet av formella krav, brist på tid och resurser samt omedvetenhet och brist på utbildning. Utifrån dessa faktorer dras slutsatser om vilka faktorer som bör beaktas vid utformningen av ett utbildningsprogram.

Studien är en kvalitativ fallstudie som utfördes i samarbete med en regional verksamhet. Studien visar att utbildningsinsatser bör vara praktiskt inriktade, kopplade till dagligt arbete och grunda sig i styrdokumenten för att möjliggöra öppna och givande diskussioner.

Understanding dynamic biological processes is difficult because these systems are continuous and emergent by nature. Traditional static or disconnected learning materials often fail to show how causes and effects unfold over time. This study examines whether a physics-based interactive simulation improves learners’ conceptual understanding of a dynamic biological process compared to a discrete step-based interactive representation. Using an exploratory comparative design, a pilot study was conducted in which participants were assigned to either a high-continuity condition (a Unity-based 3D physics simulation) or a low-continuity condition (a web-based representation using key-frame screenshots and text). Both conditions covered the same biological mechanism of the Heliamphora. Results show that both designs were highly effective: all pilot participants corrected their prior misconceptions and achieved perfect scores on the post-test. The full study (N=10), using a revised assessment instrument, revealed that the web-based condition showed a marginal advantage at systemic reasoning, while at transfer the difference between conditions was statistically and practically negligible, with only a small numerical pattern in favor of Unity. This study treats interaction continuity as a key variable and offers practical guidance for the design of interactive science education tools.

I denna studie jämfördes krypteringstid för två olika blockchifferdriftlägen, CBC och GCM för att undersöka ifall något av dem presterade bättre. Detta genomfördes på textfiler - inlägg - av varierande filstorlekar. Inläggen hade tre möjliga filstorlekar: Small (S), Medium (M) och Large (L), där varje storlek definierades av ett intervall med antal möjliga tecken. Inläggen skapades i en webbapplikation och lagrades i en databas efter kryptering för att bättre reflektera verkliga omständigheter. Vid dekryptering hämtades inläggen från databasen och dekrypterades i webbapplikationen. Kryptering- och dekrypteringstid mättes för varje filstorlek i mätserier med 3000 mätpunkter. GCM visade bättre tider i majoriteten av mätserierna, där skillnaderna blev tydligare vid större filstorlekar. Detta stämde överens med observerad bakgrundsinformation, som specificerar att GCM har möjlighet till bättre hastigheter. Resultaten gav tydligare skillnader i uppmätt tid för kryptering än dekryptering, vilket också stämde överens med flödet i krypteringsprocessen för GCM i jämförelse med CBC. Därmed drogs slutsatsen att GCM var ett generellt bättre

Emotional experiences play a central role in how players engage with and interpret digital games. Rather than merely completing objectives, players actively choose games to meet emotional needs and reflect on the outcomes of play. This thesis explores emotional immersion as a situated, player-centered experience shaped by emotional expectations, play situations, and game genres. Drawing on research in Game User Experience, player motivation, and frame theory, the study establishes a conceptual foundation for examining how emotional immersion emerges across everyday gameplay. A qualitative diary-based research method was used to capture players’ emotional states before, during, and after play, as well as reflections on whether gameplay meets emotional expectations. By focusing on lived experience and interpretation, the thesis aims to contribute a nuanced understanding of emotional immersion as a dynamic and context-dependent process.

This study investigates how proactive AI behaviours in social robots shape gaming motivation among adults aged 65 and older. Using a mixed‑methods approach, it first assessed general gaming motivations through an online survey (n = 17) and then examined how these motivations translated into human–robot interaction during playtests with a proactive gaming prototype on social robot Pepper (n = 3). Results show that core motivators—challenge, autonomy, and cognitive engagement can carry over into robot‑mediated gaming, whereas escapism and immersive solo engagement represent a structural divergence that social robots cannot replicate. Relatedness emerged as a central driver, expressed not only through human communities but also through personalised, relational interaction with the robot.

Proactive features that conveyed social presence, such as encouragement, name recognition, and adaptive difficulty, were consistently motivating, whereas context‑insensitive behaviours were perceived as intrusive. Together, these findings suggest that proactive AI behaviours hold meaningful potential as a motivational mechanism in robot-mediated gaming for older adults, particularly when designed around social presence, authentic feedback, and user-configurable proactivity. [150 words]

Det här arbetet mäter och jämför skillnader i exekveringstider mellan noisealgoritmerna Perlin-, Simplex- och Value noise, med mål att jämföra hur väl anpassade olika algoritmer är för procedurell terränggenerering i realtid.

En artefakt skapades, där ett flertal implementationer av algoritmerna testades i olika testfall och exekveringstiderna sparades för att jämföras.

Den snabbaste implementationen som mättes var en implementation av Perlin noise, men på grund av skillnader i optimeringsnivåer av olika implementationer är det oklart ifall detta berodde på att algoritmen eller endast implementationen var snabbare. För implementationer som hade samma skapare var resultatet att Value noise var snabbast, följt av Simplex noise och till sist Perlin noise, vilket stämmer mer överens med tidigare forskning.

Framtida arbeten kan utforska skillnader mellan olika noisealgoritmer i fler miljöer, exempelvis i andra programmeringsspråk eller genom exekvering på grafikkortets GPU. Studier skulle även kunna jämföra andra aspekter av noisealgoritmer, exempelvis minnesanvändning eller visuella aspekter.

Bakgrunden innehåller information av de två databaserna samt hur Node.js fungerar. Problemet till arbetet är att det saknas forskning som har analyserat skillnaden i responstid mellan MySQL och MongoDB genom användning av Node.js. Studien utfördes i en experimentell metod. Metoden bestod av att skapa två databaser och därefter koppla ihop de med Node.js. Utföra mätningar med artificiell data för att mäta responstiden hos databaserna genom Node.js. Mätningsfunktionerna testades i en pilotstudie och slutsatsen blev att ta bort det initiala värdet för att det visade en betydligt högre responstid än resterande värden. Två mätserier med olika krav utfördes och resultaten visualiserades med linjediagram, stapeldiagram och ANOVA-test. Datan analyserades och det visade sig att det fanns en signifikant skillnad där MySQL hade lägre responstid än MongoDB. Framtida arbeten kan vara att gå mer in på djupet och inte bara kolla datamängder när sökförfrågningar ställs.

Mindfulness-based interventions have gained attention as methods for enhancing emotion regulation, yet their underlying neural mechanisms remain unclear. This systematic review examined the effects of mindfulness on emotion regulation as measured by fMRI. A systematic search in Scopus and Medline EBSCO identified three empirical studies meeting all inclusion criteria. All studies used task-based fMRI with healthy adults and mindfulness-related conditions. Across studies, mindfulness conditions were associated with statistically significant reductions in negative affect and significantly modulated activity in emotion-related brain regions, particularly the amygdala. Two regulatory pathways were identified: top-down regulation involving increased prefrontal engagement, and bottom-up attenuation of subcortical reactivity without prefrontal involvement. These findings suggest that mindfulness enhances emotion regulation through complementary neural mechanisms depending on the type and duration of practice, supporting its value as an accessible intervention for improving emotional functioning.

The early caregiving environment is a key predictor of children’s brain development. Research within the field has historically focused on maltreatment or severe adversity, leaving the effects of normative parenting less explored. To address the knowledge gap and the methodological heterogeneity within the field, this thesis aims to investigate how normative variations in parenting influence children’s brain development as revealed by structural and functional neuroimaging. The systematic literature search was conducted across three databases (Medline EBSCO, Web of Science, and Scopus). A total of 1,446 records were identified, and 8 met the inclusion criteria. To determine neurodevelopmental impact, variations in parenting styles were categorized into positive and negative parenting. The findings suggest that parenting is associated with differences in brain structure and function, although the results are complex. Negative parenting was associated with changes in brain development related to accelerated maturation of amygdala-prefrontal connectivity, reduced amygdala volume, and alterations in global brain connectivity. In contrast, positive parenting was associated with larger hippocampal and total brain volume and seems to buffer the negative effects of stress. Together, variations in parenting behaviors may involve different neural circuits. Negative parenting is associated with stress-related adaptations, while positive parenting supports more typical maturation. These results highlight that even normative variations in caregiving are relevant for children’s neurodevelopment, with implications for early prevention and public health.

Background: A large part of palliative care is currently provided within municipal primary care, where care is often delivered in the patient’s home. The districtnurse has an important role and an overall responsibility to meet the needs of patients and their families.

Purpose: The aim of the study was to illuminate experiences of existential conversations with patients in municipal palliative care.

Method: A qualitative systematic literature review in which twelve scientific articles were analysed as a meta-synthesis. The search was conducted in the CINAHL and Medline databases, and the search terms were analysed using the SPICE- model.

Results: The results are presented in two main themes: An enhanced caring approach in existential conversations and Experience-based strategies for building trust, together with six sub-themes. The findings describe the complex and significant role that existential conversations play within municipal palliative care, with a focus on the approach and manner of engagement.

Conclusion: By building a trusting relationship with the patient and their family, demonstrating active listening, and being genuinely present, the conditions are created for the patient to feel safe and able to open up to existential conversation. To address the difficult questions and discussions that arise during the palliative phase, both professional experience and adequate organizational support are required.

Background: Menopause is described as a physiological process of normal aging that all women undergo. Primary health care has a responsibility to offer support, advice and treatment to women who are experiencing discomfort. Support from healthcare is described as important for maintaining good health. There has previously been limited research from the perspective of the district nurse on health promotion and care for menopausal symptoms within primary health care.

Aim: To describe district nurses' experiences of caring for and working to promote health for women experiencing menopausal symptoms in primary health care.

Method: A qualitative study was conducted with twelve district nurses in primary health care. Data collection was carried out using semi-structured interviews. The interviews were analyzed through a qualitative content analysis.

Results: The results were presented in three categories: Holistic perspective in history taking and conversation; Raise awareness, motivate and provide self-care advice; Evidence-based approach and team collaboration. The categories consist of seven subcategories.

Conclusion: The results showed that a fundamental aspect of health-promoting work for the district nurses´ was maintaining a caring perspective by attentively listening to the woman’s account of her symptoms and validating her experience of the situation. Caring conversations enabled person-centered care, as they strengthened the individual’s autonomy in performing self-care. The findings also indicated that supportive and motivational conversations from the district nurse were health-promoting, as they acted as a driving force for lifestyle changes.

Background: The number of patients wishing to be cared for at home at the end of life is increasing. For the end to be as pain free and dignified as possible, the patients’ situation needs to be viewed from a holistic perspective. District nurses are expected to coordinate the patients’ care which can create a challenging work situation which needs to be highlighted.

Purpose: The purpose of this study was to describe experiences of caring for adult patients at the end of life within home health care and in nursing homes.

Method: A qualitative method with an inductive approach was used. Data were collected through database search for scientific articles. Collected data were analyzed through metasynthesis.

Results: The results of the analysis generated four themes; Need for competence and resources, Importance of communication, Cooperation creates safety and Next of kin as a part of the care team.

Conclusion: End of life care within home health care and in nursing homes is affected in many areas by a lack of resources and competence. The work of district nurses is also affected by lack of cooperation, especially with physicians. Next of kin have a central role in the patient’s care, which can be beneficial as well as challenging. Through a person-centered approach the patient’s quality of life can be promoted throughout the palliative care process.

Pancreatic islets (also called islets of Langerhans) contain three major endocrine cell types: alpha, beta, and delta cells, which communicate through paracrine and electrical signalling to uphold blood glucose levels. Delta cells secrete somatostatin, which inhibits neighbouring alpha cells via somatostatin receptor subtype 2, suppressing glucagon secretion. Despite the recognised importance of this paracrine relationship, the spatial dependence of delta cell inhibition on alpha cells remains poorly understood. This study used optogenetics, technology that uses light to control biological processes, to investigate how delta cells influence dextran uptake in neighbouring alpha cells in mouse pancreatic islets. Dextran uptake can be used as a proxy for cellular activity in the islets of Langerhans, as it reflects endocytic and membrane trafficking processes. These processes are closely linked to metabolic activity and hormone secretion in islet cells. This study also investigated whether the physical distance between alpha and delta cells modulates this effect. ChR2-SST transgenic mice were used, in which delta cells selectively express channelrhodopsin-2 (ChR2) fused to yellow fluorescent protein (YFP). Islets were divided into groups of high/low glucose and lights on/off: Fluorescent dextran was quantified in individual alpha cells. At low glucose, optostimulation produced an approximately 8-fold reduction in alpha cell dextran uptake compared to non-stimulated controls (p < 0.0001), consistent with somatostatin-mediated paracrine inhibition via SSTR2. At high glucose, a significant but reversed effect was observed (p = 0.0010), possibly reflecting somatostatin granule depletion under supraphysiological glucose conditions. A spatial analysis revealed a trend toward lower dextran uptake in alpha cells that directly contact delta cells than in distal cells (p = 0.0939), consistent with distance-dependent paracrine inhibition. These findings support the hypothesis that delta cells exert spatially graded paracrine inhibitory control over neighbouring alpha cells. Results should be interpreted as exploratory pending validation in larger studies.

Pancreatic islets are clusters of endocrine cells that regulate blood glucose homeostasis through coordinated communication between different endocrine cell populations. Recent research suggests that spatial relationships between endocrine cells may influence β-cell activity and signalling dynamics within the islet microenvironment.

The aim of this study was to investigate dextran uptake in mouse pancreatic islets under optogenetically stimulated and non-stimulated conditions. In addition, exploratory spatial analysis was performed to investigate the proximity between β-cell-associated regions and neighbouring δ cells under different experimental conditions.

Mouse pancreatic islets were exposed to different glucose concentrations and optogenetic stimulation conditions before fluorescence imaging. Whole-islet analysis was performed on individual islets. Whole-islet analysis did not reveal statistically significant differences between experimental conditions. An exploratory ROI-level study was performed to investigate the dextran uptake within insulin-positive β-cell-associated regions. This analysis demonstrated heterogeneous dextran fluorescence patterns and revealed significant differences, however, these findings should be interpreted with caution due to pseudoreplication and the limited number of biological replicates. Exploratory spatial analysis did not reveal statistically significant differences in the proximity of β-cell-associated regions to neighbouring δ cells.

Together, the findings suggest that dextran uptake and membrane trafficking activity within pancreatic islets may display local heterogeneity. The study further highlights the potential importance of spatial endocrine cell interactions within pancreatic islets and provides a foundation for future investigations using larger biological sample sizes and automated spatial analysis approaches.

Inflammation is key biological processes involved in the development of metabolic and cardiovascular diseases. Dietary composition can influence inflammatory pathways through changes in glucose metabolism, insulin signalling, body weight and ketone production. The aim of the study was to investigate the effects of short-term low carbohydrate, high protein-high fat dietary intervention on inflammation related molecular, metabolic and physiological parameters. Four healthy participants were involved in this study, followed a high protein diet with carbohydrates restricted to 10 grams per day for 21 days. Blood samples were collected before and after the intervention. Physiological parameters, biochemical biomarkers, amino acid concentrations and gene expression level of tumor necrosis factor (TNF), nuclear factor erythroid 2 related factor 2 (NFE2L2) and sirtuin 1 (SIRT1) were analysed. Gene expression was measured using qPCR. The dietary intervention resulted in a marked reduction in carbohydrate intake from 265 g per day to 8.7 g per day, while protein intake increased from 84.1 grams per day to 148.6 grams per day. Mean fasting glucose levels decreased from 5.23 ±0.12mmol/L to 4.24 ± 0.06 mmol/L while β hydroxybutyrate concentration showed increasing trend from 0.08 to 2.00, indicating a potential induction of ketosis. Mean body weight decreased by 5.95kg after the dietary intervention. Reduction in both systolic and diastolic blood pressure were observed following dietary intervention. HDL cholesterol decreased, while LDL and total cholesterol showed slight decrease while remaining in their respective clinical reference ranges. Inflammatory markers showed mixed response however, hs-CRP concentrations decreased following the intervention suggesting potential reduction in systemic inflammation.

Overall findings suggest that high protein, low carb and high fat dietary interventions may improve several metabolic and physiological parameters associated with inflammation. The observed reduction in fasting glucose levels, body weight, blood pressure, hs-CRP indicate a trend towards a less inflammatory metabolic profile. However, the small sample size limited statistical power of this study. Further studies with larger sample size are required to confirm the effects of carbohydrate restriction on inflammation related biomarkers and gene expression.

Introduction: Children and young people are moving less, and sedentary behaviour is getting more common due to the digital transition. Digital tools can be used to promote physical activity in schoolchildren in their teens but also contribute to increasing sedentary lifestyles. Purpose: The purpose was to investigate whether and in what way digital tools can promote physical activity in schoolchildren in their teens. Method: The study was conducted as a structured literature study. The literature searches were conducted in Web of Science and Medline EBSCO. A total of 22 scientific empirical studies were included after review with CASP. The analysis was conducted according to Braun and Clark's (2006) thematic analysis. Results: The results were categorized in three themes: Digital tools that can affect physical activity, Digital feedback that can affect motivation, Use of digital tools in and outside of school. Digital tools can affect physical activity levels in schoolchildren in their teens, but to different extents and degrees. Conclusion: Digital tools can help increase physical activity among school-based delivery, especially when they are personalized and include feedback, but the results vary and should be interpreted with caution. Further research should investigate whether interventions with digital tools promote physical activity and long-term effects. Also on how user-friendly the interventions are for the target group.

Moderna webbapplikationer använder ofta API:er för att hämta och presentera data från en server. Två vanliga API-modeller är REST och GraphQL, vilka skiljer sig åt i hur data exponeras och hämtas. När data innehåller många relationer kan valet av API-modell därför påverka svarstiden. Syftet med denna studie var att undersöka hur REST och GraphQL skiljer sig i svarstid vid hämtning av data från en relationsdatabas med olika nivåer av relationsdjup. Studien genomfördes som ett kontrollerat experiment där båda APImodellerna implementerades i samma lokala testmiljö med Node.js, PostgreSQL och SimpleEnglish Wikipedia-data. Tre testfall användes: noll relationsdjup, lågt relationsdjup och högt relationsdjup. För varje testfall genomfördes upprepade mätningar med ett automatiserat mätskript. Resultaten visade att REST hade lägre svarstider än GraphQL vid noll och lågt relationsdjup. Vid högt relationsdjup förändrades däremot resultatet tydligt, då GraphQL hade betydligt lägre svarstider än REST. Slutsatsen är att valet mellan REST och GraphQL bör göras utifrån datans struktur och hur många relationer som behöver följas. REST kan vara mer effektivt vid enklare hämtningar, medan GraphQL kan vara mer fördelaktigt vid hämtning av relationsrik data med högre relationsdjup.

Denna studie undersökte Document Object Model (DOM)-uppdateringstid mellan React.js och Svelte.js i en Single-Page Application (SPA), där React.js använde en virtuell DOM och Svelte.js manipulerade DOM direkt med genererad JavaScript. Problemet var att snabba DOM-uppdateringstider kunde påverka användarupplevelsen, därför var DOM-uppdateringstid en relevant mätvariabel. Ett experiment genomfördes där två applikationer utvecklades, en i React.js och en i Svelte.js, och DOM-uppdateringstiden mättes när data lades till i en tabell med 50, 500 och 5000 mätpunkter. Ett mätskript skrevs och användes med Tampermonkey ett tillägg för webbläsare. Varje kombination av ramverk och antal mätpunkter itererades 3 gånger. Resultatet visade att vid 50 mätpunkter fanns det ingen statistisk signifikans, medan statistisk signifikans fanns för 500 och 5000 mätpunkter där Svelte.js presterade snabbare än React.js. I framtida arbete skulle bland annat fler iterationer utföras, fler CRUD-operationer skulle användas och fler webbläsare skulle ha mätts på.

Denna studie vill undersöka skillnader i prestanda i grafik-API:erna WebGL och WebGPU med användning av de ledande tredjepartsbiblioteken för 3D-grafik i webbläsaren; Three.js och Babylon.js. Prestandan som mäts är Frames per Second (FPS) och renderingstid för varje frame. En spelsimulation har tagits fram i fyra olika varianter och mätts med hjälp av mätscript i ett experiment. Vad som hittades i resultaten var att WebGPU presterar bättre än WebGL i renderingstid, men att WebGL fortfarande är överlägsen WebGPU i FPS. Det som påverkar prestandan mest verkar vara valet av tredjepartsbibliotek. Babylon.js presterar dock likvärdigt med båda grafik-API:er som renderare. Tidigare studier har visat att WebGPU presterar bättre än WebGL i flera olika sammanhang, men behöver undersökas ytterligare för att kunna förstå vilka specifika kontexter de båda API:erna presterar bäst i, och vad WebGPU saknar för att helt kunna ta över från dagens standard, WebGL.

Chemotherapy-induced intestinal mucositis is a common complication of cancer treatment, characterised by epithelial damage, inflammation, and impaired nutrient absorption. While proglucagon-derived peptides such as GLP-1 and GLP-2 are known to support intestinal integrity, the role of glucagon in intestinal inflammation remains unclear. This study aimed to investigate changes in glucagon levels during acute 5-fluorouracil (5-FU)-induced intestinal mucositis in mice and to assess how different analytical methods influence glucagon measurement. Plasma and intestinal glucagon levels were measured using a radioimmunoassay (RIA) and a sandwich enzyme-linked immunosorbent assay (ELISA). Intestinal tissue extracts from the duodenum, jejunum, and ileum were prepared using acid-based or phosphate-buffered saline extraction. Glucagon concentrations from intestinal tissue extracts were normalised to total protein content as determined using a bicinchoninic assay. Plasma glucagon levels were significantly increased in 5-FU-treated mice compared to controls. RIA measurements showed significantly elevated glucagon concentrations across all intestinal regions in 5-FU-treated mice compared to controls, with the greatest increase observed in the ileum. In contrast, ELISA measurements were more variable, with significant increases primarily observed in the RIA-extracts from the ileum and PBS-extracts from the jejunum and ileum of the 5-FU-treated mice compared to the controls. In summary, the results show that glucagon levels are increased during acute intestinal mucositis, systemically and regionally; however, the magnitude and consistency of this increase depend on the analytical approach used.

Detta examensarbete undersöker hur valet av animationsmetod påverkar prestanda i webbläsarbaserade spel. Studien utgår från tidigare forskning om webbanimationer, rendering och animationsbibliotek, där det framgår att olika tekniker kan ge olika prestandautfall. För att undersöka detta utvecklades två versioner av ett match-3 spel. En med manuellt implementerade animationer i JavaScript/CSS och en med animationsbiblioteket GSAP. Båda versionerna testades med samma spelmekanik och ett eget testverktyg som mätte FPS, minnesanvändning och LongTasks. Resultaten visade statistiskt signifikanta skillnader mellan versionerna. GSAP-versionen uppvisade lägre minnesanvändning, färre LongTasks och något stabilare FPS. Studien visar att animationsmetod kan påverka prestanda, men resultaten bör tolkas utifrån den specifika spelmiljön. Framtida arbete bör omfatta fler speltyper, längre tester och fler webbläsare.

Denna studie undersöker renderingstiden i vektorbaserade interaktiva webbkartor genom att jämföra Leaflet.VectorGrid och MapLibre GL JS. Experimentet fokuserar på att mäta renderingstid efter interaktioner genom att använda kamera-event som en proxy för användarupplevd interaktionslatens. Flertalet scenarion har genomförts i en kontrollerad miljö för att utvärdera skillnader i renderingstid mellan de två biblioteken. Resultatet visar att MapLibre GL JS är konsekvent snabbare än Leaflet.VectorGrid, men proxyn som använts är inte den mest representativa för slutförd visuell rendering. Caching-inställningar kan påverka Leaflet.VectorGrid negativt, medan MapLibre GL JS inte visar någon skillnad. Även geografisk komplexitet kan påverka renderingstiden i Leaflet.Vectorgrid om cache är aktivt, men i övriga sammanhang finns ingen statistiskt signifikant grund för detta. Fortsatt arbete är möjligt genom andra tekniker, enheter, eller mängd data som ska renderas.

Denna studie undersöker prestandaskillnader mellan C++ och Rust via WebAssembly för 3D-grafik i webbläsaren med Three.js. Bakgrunden är att JavaScript har begränsningar vid beräkningstunga operationer, vilket gör WebAssembly intressant för partikelsystem, och att det inte finns mycket forskning om vilket språk som är mest effektivt när man använder C++ och Rust med Three.js via WebAssembly. Syftet är att jämföra exekveringstid, FPS och minnesanvändning mellan C++ och Rust. Ett partikelsystem implementerades med Three.js, där C++ kompilerades med Emscripten och Rust med wasm-pack/bindgen till WebAssembly. Mätningar genomfördes i Chrome och Firefox med performance.now för FPS och exekveringstid samt DevTools för minnesanvändning. Resultatet visar statistiskt signifikanta skillnader mellan språken i FPS, exekveringstid och total minnesanvändning, medan heapminne utan arraybuffertstorlek inte visade någon signifikant skillnad. Framtida arbete kan undersöka skalning, rotation, CPU-belastning, GPU-beräkningar och fler optimeringsnivåer samt hur tekniken kan vidareutvecklas för mer avancerade simuleringar.

As a factor affecting convolutional neural network (CNN) classification performance, hyperparameters require optimization. The relative efficacy of hyperparameter optimization algorithms is task-dependent, but has not been established for CNNs fine-tuned on agricultural pest datasets. Accordingly, this study evaluates the impact of Random Search (RS) and Bayesian Optimization (BO) on CNN classification performance across three visually distinct pest domains: plant diseases, agricultural weeds, and crop pests. Three quasi-experiments were conducted, each employing RS and BO separately to optimize the hyperparameters of MobileNetV2 when fine-tuned on a specific dataset (ASDID, DeepWeeds, or IP41). The models trained with BO-optimized hyperparameters achieved significantly higher test set performance on DeepWeeds and IP41 than those optimized with RS, whereas no significant difference was observed for ASDID. These results suggest that RS and BO have comparable efficacy for a dataset with lower visual and structural complexity, whereas the advantage of BO scales with increasing dataset complexity.

This study examines how different AI-based document processing paradigms perform in automated information extraction from heterogeneous transactional documents. Using a quantitative quasi-experimental design, three paradigms are evaluated: intelligent document processing(IDP) systems, general-purpose multimodal large language models, and local large language model pipelines with preprocessing. The systems are tested on transactional documents divided into three datasets by document type: order confirmations, invoices, and receipts. Extraction performance is evaluated primarily in terms of accuracy, while precision, recall and F1-score are also taken into account.The findings show that extraction performance depends not only on the selected system, but also on document type, layout variation, and preprocessing conditions. Overall, the IDP paradigm showed the strongest extraction accuracy when document layouts were known and represented in prelabelled training data. The general-purpose LLM paradigm proved the most flexible solution, while the local LLM paradigm showed the weakest performance in general. The significance of the results is supported by confidence interval analysis and statistical tests. It is concluded that no single paradigm is superior across all extraction tasks, and that the suitability of a solution depends on the operational context. By examining how different document processing paradigms perform under varying document conditions, this study contributes to the development of more reliable and practically applicable solutions for automated information extraction from transactional documents. Future work could examine larger or alternative local models, more targeted prompting strategies, varying amounts of labelled training data, or how retrieval-augmented generation (RAG) might further improve extraction performance.

Condition monitoring analysts routinely interpret vibration graphs and translate their findings into structured assessment fields in maintenance reports. This is a repetitive process that depends on consistent expert judgement. The aim of this thesis is to investigate whether a vision-language model can support this step by predicting four target fields directly from graph images. Failure Cause, Diagnosis Confidence Level, Failure Probability Factor and Action Priority. A zero-shot prompted baseline is compared against a supervised vision fine-tuned GPT-4.1 trained on historical graph-label pairs, evaluated using five-fold GroupKFold cross-validation on 209 samples drawn from 34 maintenance reports. The baseline performs reasonably on the numeric fields but fails almost entirely on the categorical ones. Hence showing the difficulty of applying domain-specific reporting conventions without training data. Fine-tuning produces substantial improvements across all four fields, with Action Priority rising from near-zero to above 80% accuracy and both numeric fields reaching near-perfect scores, suggesting that supervised vision fine-tuning on a small labelled dataset is a viable path toward AI-assisted maintenance reporting.

Many diagnostic tools for colorectal cancer are currently available for screening and diagnosis. However, detection often occurs at later stages, when metastatic lesions have formed. Analysis of genes through DNA methylation alterations in polyps may suggest malignant potential of a polyp, thereby allowing for early detection. Using peripheral blood mononuclear cells (PBMC) as a source of DNA reduces the invasiveness associated with analysis through formalin-fixed paraffin-embedded (FFPE) tissue. This study aimed to identify DNA methylation differences in LINE1, a global methylation marker, and hMLH1, a key component of the mismatch repair system, between PBMC from healthy controls and PBMC and FFPE polyp tissue from patients with potentially malignant colon polyps. DNA methylation differences were assessed through bisulphite pyrosequencing. hMLH1 did not show differences in the majority of CpG sites. The LINE1 methylation level was significantly different between polyp tissue (n = 47) and PBMC from controls (n = 46) at all CpG sites (CpG site 1: p < 0.001; CpG site 2: p = 0.004; CpG site 3: p = 0.002). Two out of three CpG sites showed a statistical difference between PBMC from the polyp (n = 54) and control (n = 46) groups (CpG site 1: p = 0.013; CpG site 2: p = 0.020). While FFPE appeared to be the superior source for DNA methylation analysis, PBMC serves as a potential contender through site-specific analysis. Analysis of additional CpG sites for LINE1 is suggested for further validation of LINE1 as a biomarker with PBMC as a source.

We study path rings, Cohn path rings, and Leavitt path rings associated to directed graphs, with coefficients in an arbitrary ring R. For each of these types of rings, we stipulate conditions on the graph that are necessary and sufficient to ensure that the ring satisfies either the rank condition or the strong rank condition whenever R enjoys the same property. In addition, we apply our result for path rings and the strong rank condition to characterize the graphs that give rise to amenable path algebras and exhaustively amenable path algebras.

Background: Primary healthcare (PHC) is often a person’s first contact with the healthcare system and they often meet district nurses. The need and demand for digital care encounters are increasing in healthcare services. The aim of this study was to highlight district nurses' experiences of working with digital care encounters in PHC.

Methods: In this pilot study, a qualitative descriptive design was used to gain a deeper understanding of district nurses’ experiences. Data were analysed using qualitative content analysis according to Graneheim and Lundman [1].

Results: The analysis resulted in one overarching theme and four categories. District nurses in PHC experienced digital patient encounters as complex, captured in the theme “complexity in navigating digital care in PHC.”, Four categories were identified: “digital care encounters entail opportunities”; “obstacles in digital care encounters”; “the impact of digital interaction on the patient encounter”; and “expectations and development needs in digital PHC.”

Conclusion: Digital care encounters represent a new and complex aspect of district nursing that should be considered in implementation. Clearer routines and guidelines may support both nurses and patients in optimising digital care.

Background: Dreaming during anaesthesia has been documented for over a century, yet it remains poorly understood. This scoping review synthesises evidence on the frequency, phenomenology, predictors, outcomes, and neural correlates of anaesthesia dreaming.

Methods: We systematically searched multiple databases for studies examining anaesthesia dreaming, including clinical trials, observational studies, and experimental investigations in adult and paediatric patients, as well as in healthy volunteers.

Results: We identified 157 studies (1921–2024) involving 87 866 participants. Dream recall varied widely, with higher rates in experimental settings (often 40%–80%) compared with large-scale clinical trials (typically 3%–8%), likely reflecting methodological differences. Ketamine (clinical: 0%–95%; experimental: 44%–100%) and propofol (clinical: 0%–64%; experimental: 0%–77%) were most consistently associated with recall. Dreams were predominantly pleasant, featuring everyday autobiographical content. Higher home dream recall frequency, preoperative suggestion or priming, and immediate post-emergence assessment were associated with higher recall. Evidence regarding clinical outcomes was limited, but unpleasant dreams were linked to lower willingness to undergo the same anaesthetic again. Dream recall was not associated with intraoperative awareness, challenging interpretations of dreams as markers of inadequate anaesthetic depth. Emerging neural evidence suggests that dreaming exhibits signatures distinct from unconsciousness and connected consciousness.

Conclusions: Anaesthesia dreaming is common, predominantly positive, and pharmacologically specific. Major gaps in knowledge include inconsistent definitions, inadequate assessment methods, limited mechanistic understanding, sparse paediatric research, and minimal outcome data. Standardised approaches informed by consciousness science and sleep research are needed. The positive nature and suggestibility of anaesthesia dreams suggest potential therapeutic applications warranting investigation.

Introduction: Patient-accessible electronic health records (PAEHRs) offer benefits, such as supporting self-management and care engagement. However, some patients, particularly those with mental health conditions, might experience negative emotions such as worry when reading unexpected or sensitive information in their PAEHRs.

Methods: A web-based survey of 4459 respondents distributed via the Finnish national patient portal included multiple-choice and open-ended questions. Respondents consisted of two patient groups who had received care either for 1) mental health or 2) other conditions. Inductive content analysis was performed to explore the kind of information that was perceived as sensitive in the PAEHR. Associations between sociodemographic factors including the type of care and reporting health information as sensitive were calculated via the multivariable binary logistic regression analysis.

Results: Mental health (61.3%), and intimate health (8.3%) were the most frequently mentioned as especially sensitive types of information among respondents, who also stressed that the sensitive nature of the health information depended on the context. Within the mental health information type, therapy or treatment was most often mentioned (3.5%) as sensitive. Respondents who had received mental health care were significantly more likely to perceive certain information as sensitive (53.2%) than other patients (28.8%; Adjusted OR=2.783, 95% CI=[2.333, 3.319], p<0.001).

Conclusions: This study delves into the sensitive character of mental health information within PAEHR. The sensitivity of information also depends on the consequences for the patients when data will be used in another context. Documenting sensitive information carefully and safeguarding it is recommended to maintain trust in electronic health records and healthcare. 

Introduction: Social media is a central part of children and young people's everyday lives and is used daily by the majority of children and young people. Research shows that social media affects mental health both positively and negatively. Despite extensive research, the results are partly contradictory, which is motivated by investigating how children, young people and related adults experience social media as being affected by mental health. Purpose: The purpose was to investigate young people and related adults' experiences of risk and protective factors linked to young people's use of social media. Method: The study is conducted as a structured literature study using a qualitative method. 21 qualitative original articles were included and analyzed through thematic analysis. Results: Both children, young people and related adults described that social media affects children and young people's health in several ways. They highlight risks such as social comparison, sleep difficulties and mental illness, but also positive aspects such as social support, community and the opportunity for self-expression. The results show that social media can function as both a risk factor and a protective factor for children and young people's mental health. Conclusion: The study shows how children, young people and related adults experience that social media has both positive and negative effects on children and young people's mental health. The impact depends on usage patterns, individual circumstances and social contexts. It is important to promote conscious and healthy use of social media among children and young people.

This study examines how individuals in leading roles within large organizations design and use governance mechanisms in working with artificial intelligence. The background of the study is the increasing use of AI in organizations, where the technology is increasingly applied as support for analysis, planning, and decision-making. This development has changed the conditions for responsibility, control, and follow-up, while the complexity and limited transparency of AI systems create new challenges. Despite the existence of established theoretical frameworks for AI governance, there is a lack of empirical knowledge about how these are applied in practice, particularly by leaders. 

The purpose of the study is to investigate how individuals in leading roles design, adapt, and apply governance mechanisms in AI-related work. To address this, a qualitative method has been used, where eight semi-structured interviews were conducted with individuals in leading AI-related roles within both the public and private sectors. The collected material has been analyzed using a theoretical framework that includes a sociotechnical perspective, sensemaking, and theories of AI governance. 

The findings show that AI governance in organizations is carried out through a combination of structural, procedural, and relational mechanisms. Structural mechanisms, such as policies and guidelines, provide frameworks for how AI may be used. Procedural mechanisms include, for example, review, monitoring, and quality assurance of AI systems. Relational mechanisms, such as communication and collaboration between different functions, are shown to be crucial for enabling AI to function effectively in practice. 

The study also shows that AI transforms decision-making by becoming an active part of the process, which affects both the distribution of responsibility and the degree of control. In particular, AI’s autonomy and limited transparency create challenges for leaders, as it can be difficult to understand how decisions are generated. Trust and understanding of AI emerge as central factors influencing how systems are used and how much decision-making authority is delegated to the technology. 

The conclusion is that effective AI-supported decision-making requires a balance between human judgment and AI-generated recommendations. AI governance should therefore be viewed as a dynamic practice, where formal structures are combined with social and organizational processes to manage responsibility, control, and technological autonomy in an increasingly AI-driven context.

AI-baserade informationssystem som bygger på hämtningsförstärkt generation (RAG) har blivit en lösning för att tillgängliggöra information i organisationer, då de kompletterar stora språkmodeller med extern, icke-parametrisk kunskap. Ur ett informationssystemsperspektiv är hämtningssteget avgörande eftersom det styr vilken information som identifieras och tillgängliggörs för användaren. Relevansen hos den hämtade informationen kan därmed förstås som en dimension av informationskvalitet, vilket i sin tur utgör en framgångsdimension för informationssystemet. För att stärka denna relevans har flera designstrategier utvecklats, däribland HyDE, HyPE och hybridsökning. Tidigare forskning behandlar dessa främst som isolerade tekniska optimeringar.

Studien antar ett designvetenskapligt perspektiv och utvärderar de tre designstrategierna mot en baslinje i form av tät hämtning. Artefakten realiseras som en gemensam hämtningsarkitektur där samtliga designstrategier implementeras enskilt under samma experimentella förutsättningar. Utvärderingen genomförs över fem domänspecifika dataset från BEIR-ramverket med nDCG@10 och Recall@100 som mått på artefaktens måluppfyllelse.

Resultaten visar att studiens baslinje baserad på en modern inbäddningsmodell, uppnår starka värden i samtliga domäner och i flera fall överträffar designstrategierna. HyDE och hybridsökning uppnår signifikanta förbättringar endast på SciFact (Vetenskaplig faktagranskning). HyPE uppvisade signifikant lägre relevans i samtliga undersökta domäner. Fynden indikerar att RAG baserade informationssystem är domänberoende, och att nyttan av ytterligare designstrategier kan minska när baslinjen redan uppvisar en stark hämtningsprestanda.

Denna studie undersöker hur valet av databashanterare påverkar responstiden i ett kommunalt webbaserat bokningssystem. Syftet var att jämföra MySQL och MongoDB vid historik-, sök- och bokningsoperationer under olika datamängder. Två versioner av samma bokningssystem implementerades där endast databashanteraren skiljdes åt. Systemet testades genom automatiserade användarsessioner vid datamängderna 500, 1500 och 3000 bokningar.

Resultaten visade att MySQL uppvisade generellt lägre responstid vid historik- och sökoperationer, medan MongoDB uppvisade lägre responstid vid bokningsoperationerna. ANOVA-testerna visade dessutom att skillnaderna mellan databashanterare var statistiskt signifikanta. Studien visar därmed att valet av databashanterare påverkar responstiden beroende på operationstyp och belastningsscenario. Framtida arbeten kan undersöka hybridlösningar som kombinerar effektiv hantering av både läs- och skrivoperationer.

This research paper explores alternative applications for Large Language Models (LLMs) beyond traditional text and code generation. An artifact was developed to perform pathfinding within a localized street map, capable of utilizing various algorithms. The study focuses on comparing Bidirectional Dijkstra with LLMA*, a novel approach proposed by Meng et al. that leverages an LLM to identify checkpoints, which are subsequently connected using an A* algorithm. This methodology aims to exploit the efficiency of A* in restricted search spaces while extending its reach over larger geographical areas. The experiment involved 100 iterations for each of 10 randomly selected node pairs. Performance was evaluated based on execution time, node exploration count, CPU and RAM utilization, and path length.

The results indicate that LLMA* is significantly slower than Bidirectional Dijkstra and produces suboptimal paths if any.

A research gap has been identified regarding the possible connections between player typologies and player preferences for elements of game design. This thesis investigates whether correlations exist between players’ Hexad types and their preferences for game elements. It also compares the Hexad profiles of Skövde University game development students from 2022 to today. Game elements were constructed through a thematic analysis of Steam tags, and a qualitative survey was used to collect data regarding the participants’ Hexad profiles and preferences. Analysis of the results yielded that Hexad types correlate slightly with certain types of elements. Socializers favoured multiplayer playing styles, Achievers favoured game elements and slightly tended towards playing styles, Players primarily favoured game elements, while Philanthropists and Free Spirits favoured game elements and dramatic elements alike. Disruptors were barely present in the sample. Moreover, the Hexad profiles of game development students have remained mostly stable.

Swedish municipal social services manage some of the most sensitive personal data in the public sector. As digitalisation has expanded, social workers are expected to protect client information in line with multiple overlapping legal frameworks, including GDPR, the Social Services Act (SoL), and the Public Access to Information and Secrecy Act (OSL). Despite this, little research has explored how information security is experienced and practiced by social workers in their daily work.This study explores how social workers in Swedish municipalities understand and navigate information security requirements. A total of ten semi-structured interviews were conducted with social workers from different Swedish municipalities. The data were analysed using inductive thematic analysis following Braun and Clarke (2006).Four themes were identified. First, most participants understood information security as part of their professional identity and responsibility toward clients. Second, legal frameworks shaped daily work through habits and routines rather than direct knowledge of the law, with OSL causing the most uncertainty. Third, formal training was limited or absent for most participants, and knowledge was built through experience and colleagues. Fourth, management support, clear routines, and functional digital systems made a positive difference where they existed, while their absence left some participants without reliable guidance.The findings suggest that information security in Swedish municipal social services depends heavily on individual motivation rather than consistent organizational support. This creates uneven conditions across workplaces, which is a concern given the sensitivity of the information involved.

Video games are considered interactive media in which game users or players actively engage with the system to create meaningful play. As in other medias, in order to effectively communicate to the audience, it requires a system that elaborate on the combination of languages, audio and especially visual design. While visual design in video games is often mentioned by the user interface or level design, the visual cues can also be embedded in the feedback loops. They function as a system that informs player actions, help reinforce the players heuristic, and guide decision making toward game objectives. This type of semiotic system is designed by developers to convey information in a subtle way, requiring players to interpret meanings through learning cycles in the game.

Artificial intelligence has developed rapidly in recent years, and one important development within this area is the rise of Large Language Models (LLMs). Previous research shows that LLMs may be useful in several cybersecurity and network security tasks, such as threat detection, anomaly detection, incident response, malware analysis, and vulnerability assessment. However, much of the existing research focuses on specific model performance, technical capabilities or general cybersecurity use cases, while less attention has been given to how LLMs may be understood and used in specific company settings such as fintech companies.The aim of this study was to explore how professionals in fintech companies perceive and describe the use of LLM technology for network security, including current use, potential benefits, opportunities, challenges and risks. The study used a qualitative method based on semi-structured interviews with five professionals working in fintech environments. The collected data was analyzed using thematic analysis.The results show that current use of LLMs for network security within the studied fintech companies is limited and not yet fully established. The majority of participants described no direct use but they also described early testing or using them in other ways to help with cybersecurity related work. The main potential use cases identified were code review, threat modelling, information search, monitoring, and analyzing logs or processing large amounts of data. The findings also show that LLMs are mainly viewed as support tools that can help reduce manual work in complex security environments. At the same time, several challenges were identified, including hallucinations, false positives, incorrect actions, data protection concerns, regulatory requirements, and the risk of giving AI systems too much decision making power. Human controls like human in the loop, review and trust but verify, and accountability were therefore identified as important conditions for using LLMs in network security work. Overall, the study concludes that LLMs have potential to support network security within fintech companies, but that potential implementation is shaped by practical, legal, organizational, and security requirements.

The financial gains associated with ransomware have made it a popular and serious cybersecurity threat, while organizations may have become more vulnerable to this type of attack due to their dependence on digital infrastructure. Ransomware’s association with phishing as a commonly used attack vector, which may require organizations to consider employee awareness training alongside technical security measures. Small and medium-sized enterprises (SMEs) can be more vulnerable to ransomware because of resource constraints and limited cybersecurity knowledge. That is why this thesis aims to investigate how cybersecurity awareness training can help SMEs in Europe to support employees in identifying and responding to phishing attempts in the context of ransomware risk. This was done through a systematic literature review, where previous research was retrieved from five different databases. This was done with a specified search term to find articles used to answer this thesis research question. The literature selection process resulted in 35 articles being included in the final review and analyzed with thematic coding. Findings from the analysis suggest that cybersecurity awareness training may help SMEs to support employees to identify and respond to phishing attempts. However, the findings suggest that awareness training may be more useful for SMEs when it is practical, continuous, and connected to employees’ daily tasks, reporting procedures, and incident response routines.

Denna studie undersöker vilken grad av handholding som bidrar till en tillfredställande spelupplevelse i 3D-spel. En enkel Unity-build skapades med fem nivåer där fem olika navigationssignaler (UI, ljus, färg, form och storlek) successivt trappades ner. Via en mixad metoddesign som består av semistrukturerade observationsstudier och semistrukturerade intervjuer testades 13 deltagare.

Resultaten visar att minskad handholding ledde till ökad genomsnittlig tid, frustration och systematiskt rörelsemönster hos deltagarna. Nivå 2, samt navigationssignalen ljus, ansågs bidra med mest tillfredställande grad handholding. Anledningen varför var att de balanserade personlig autonomi i spelet med tydlig vägledning. Navigationssignalen formskillnad missades helt, vilket påverkade resultatet i nivå 4 och 5 genom att deltagare började testa varje tält.

Studien visar att vana spelare föredrar naturligt integrerade navigationssignaler och att handholding är starkt kontextberoende. En del metodologiska brister ger dock resultatet en lägre trovärdighet.

Detta arbete undersöker vilka element och spelmekaniker som bidrar till game feel i 2D-plattformsspel, som presenteras av akademiska och spelindustriella källor. På grund av att kunskap angående game feel är utspritt bland dessa källtyper så har denna studie genomfört en strukturerad litteraturöversikt tillsammans med datatriangulering, strukturerat kring Pichlmair och Johansens (2022) ramverk om game feel design som delas in i de tre domänerna av fysikalitet, amplifikation och stöd. Resultaten av detta arbete indikerar att game feel inom 2D-plattformsspel inte skapas av isolerade mekaniker, utan snarare genom samspelet mellan noga och avsiktligt justerade element såsom spelarrörelse, gravitation, hopp, kollision, återkoppling, inmatningsassistans och kamerabeteende. Dessa element bidrar till att forma spelupplevelsen genom att få spelet att kännas mer responsivt, läsbart, rättvist och tillfredsställande att spela. Framtida arbete skulle kunna stärka områden med bristande stöd, test utvalda spelmekaniker empiriskt, samt applicera en mer systematisk metod.

This study examines how extremist ideologies are expressed within online video game communities and to what extent such expressions are framed as joke or humor. Drawing on contemporary research on online extremism, gaming cultures and digital radicalisation, while conceptualizing gaming communities as interactive social environments where ideological discourse may emerge, circulate and become normalised through everyday interaction rather than through direct recruitment alone. The study situates this discussion within broader debates surrounding digital extremism, by using the 2019 Christchurch mosque shootings and the 2014 Gamergate controversy as examples to showcase different forms of extremism. These events illustrate how gaming related cultures, online anonymity and platform dynamics can intersect with harassment, exclusionary ideologies and the normalisation of harmful discourse.