SACM - United Kingdom
Permanent URI for this collectionhttps://drepo.sdl.edu.sa/handle/20.500.14154/9667
Browse
71 results
Search Results
Item Unknown Engineering and Classifying Indirect Prompt Injections: A Socio-Technical Approach to AI Agent Security(Saudi Digital Library, 2026) Aldurayhim, Abdulaziz; Abudayeh, MohammadAs autonomous AI agents become more prevalent in enterprise settings, they are now susceptible to "Indirect Prompt Injection (IPI)" attacks, which embed adversarial payloads in external data sources to manipulate the agent's contextual execution and force unauthorised actions. This dissertation holds that IPI attacks constitute a socio-technical phenomenon because they exploit semantic vulnerabilities in agent architectures and psychological manipulation patterns in human communication. To address this threat, the study proposes a three-dimensional taxonomy based on payload type, injection vector, and social engineering trigger. The initial corpus comprised 1,119 instances; following rigorous preprocessing and deduplication, a refined dataset of 858 validated samples was established for modelling. A separate 1,032-sample verified dataset pack was used only as an independent reference for dataset verification, not as the definitive modelling dataset. A DeBERTa-v3-base classifier is fine-tuned to classify social engineering triggers, using RoBERTa-base as a baseline. On the OSINT subset, DeBERTa's macro-F1 score was 0.983, and the accuracy was 91.7%. The study also presents a probabilistic Confused Deputy threat model and an ACE defence architecture consisting of architectural controls, runtime detection, privilege minimisation, human-in-the-loop oversight, and practitioner education. The findings conclusively demonstrate that integrating advanced NLP detection mechanisms with socio-technical security frameworks significantly enhances enterprise resilience against IPI threats.12 0Item Unknown Essays on Artificial Intelligence and Firm Performance(Saudi Digital Library, 2026) Obaidan, Ali; Banerjee, Sourindra; Heinberg, Martin; Katsikeas, CostasArtificial Intelligence (AI) is widely regarded as the latest General-Purpose Technology (GPT) to emerge. This characterisation places AI alongside transformative innovations like the steam engine, electricity, and the internet. If such characterisation is correct, AI adoption is likely to have significant implications for firm performance. For instance, AI can enhance firms’ ability to innovate and commercialise innovations while improving productivity through reduced labour intensity and operational disruptions. However, as with prior GPTs, realising AI’s full potential requires complementary strategic and organisational adjustments that are inherently lengthy, uncertain, and costly. I assess this theory through two complementary studies. In the first, I investigate whether firms’ strategic focus on revenue growth versus cost efficiency with AI leads to differential performance outcomes. To do so, I analyse over 2 million AI-related contextual windows from firms’ annual reports to infer their AI strategic focus. My results show that focusing on cost efficiency with AI improves firm performance whereas focusing on revenue growth negatively affects performance outcomes. Moreover, I find that marketing and operations capabilities positively moderate the effect of AI revenue focus and negatively moderate the effect of AI cost focus. In the second study, I investigate the effects of hiring AI-skilled employees on firms’ innovation and marketing capabilities. In doing so, I draw on a unique dataset tracking the presence of AI-skilled employees in U.S. firms through textual analysis of their resumes. The results indicate that hiring AI-skilled employees enhances firm performance through its positive impact on innovation capability while simultaneously undermining performance through its negative effect on marketing capability. Moreover, the analysis reveals that financial slack plays a critical role in positively moderating these effects. Overall, my findings suggest AI’s impact on firm performance depends on various complementary resource investments and functional capabilities, making it far more complex than is commonly assumed.8 0Item Unknown Designing a Customer Experience Framework to Optimise the Impact of AI Adoption in the Retail Sector(Saudi Digital Library, 2026) Alenazi, Mohammed; Yinshan, Tang; Gulliver, StephenThe use of artificial intelligence (AI) in retailing is widely recognised as a significant catalyst for operational, supply chain and consumer behavioural change. However, despite substantial investment in AI, several retail organisations are struggling to achieve the desired customer experience and firm performance. While extant literature explores AI adoption or customer experience, very few studies discuss the relationship between backend operational efficiency enabled by AI and frontend customer experience. The aim of this study is to address this research gap by proposing an overarching framework of customer experience that maximises the effect of AI adoption in the retail sector. This study adopts the socio-technical perspective and an extended Leavitt Diamond Model to conceptualise AI adoption as a change agent that affects four interconnected organisational elements, namely, people, task, technology, and structure. A mixed methods approach is employed for data collection and analysis. The quantitative survey data from 390 retail employees and consumers were analysed using SPSS and Partial Least Squares Structural Equation Modelling (PLS-SEM) to examine the inter-relationships between AI adoption and some organisational and customer-related constructs. In addition, semi-structured interviews with retail managers and employees were conducted to gather insights into AI adoption implementation, operational challenges and customer experience. The reflexive thematic analysis is employed to analyse the qualitative data and triangulate with the quantitative data. The results suggest that AI adoption has a significant impact on customer purchasing behaviour, inventory management efficiency, employee productivity, shopping and task efficiency and customer experience. However, implementation, data integration and security are major constraints to successful AI adoption. Moreover, this study confirms that backend operational efficiency has a positive relationship with customer experience, which suggests that organisational efficiency plays a crucial role in enhancing customer experience. This study contributes to the extant literature primarily by extending the Leavitt Diamond Model within the context of AI adoption in the retail sector. The findings provide empirical evidence linking v AI-enabled operational efficiency to customer experience and demonstrate how the interdependent dimensions of structure, people, technology, and tasks shape AI-enabled organisational transformation. In addition, the study develops an evidence-informed practical AI adoption framework to support retail organisations in implementing AI initiatives that enhance both operational performance and customer experience. The study suggests that customer experience should be configured as an integrated framework of interdependent constructs. The framework can be used by practitioners as a tool to evaluate their readiness for AI adoption and manage the challenges associated with AI adoption. Moreover, the framework will help retail decision makers to ensure that AI adoption contributes to sustainable value for both organisations and their customers. This study argues that AI adoption should be considered as an organisation-wide initiative rather than a technology only solution. Future studies may investigate the longitudinal effect of AI adoption, governance and ethical issues in AI-enabled retailing and apply the proposed framework in other service industries.17 0Item Unknown Artificial intelligence for bias detection in higher education online content(Saudi Digital Library, 2026) Bin Shiha, Rawan; Eric, Atwell; Noorhan, AbbasThis thesis develops and evaluates Artificial Intelligence (AI) and Natural Language Processing (NLP) approaches for detecting bias in higher education online resources, addressing the lack of systematic computational methods in this area. Bias in higher education content shapes knowledge production, representation, and equity, making its detection both academically and socially significant. To address this gap, three sets of novel datasets were created: 1. a corpus of university news articles annotated for subjectivity, sentiment, and gender representation; 2. three university reading list datasets with demographic annotations enabling comparative analysis across Western and Middle Eastern contexts; and 3. two domain-specific collections of learning materials, one from humanities-oriented open resources and the other from Science, Technology, Engineering, and Mathematics (STEM) lecture transcripts. Together, these datasets provide the first systematic resources for investigating representational, stereotypical and linguistic bias across diverse higher education domains. Using these resources, a range of Pre-trained Language Models (PLMs) and Large Language Models (LLMs) were evaluated for bias detection. PLM revealed significant gendered and representational disparities in university discourse, while fine-tuned LLM achieved improved performance on humanities data but showed limited transferability to STEM materials. A hybrid framework integrating fine-tuned LLMs with Retrieval-Augmented Generation (RAG) enhanced detection transparency and prediction balance. These methods were operationalised in a prototype web application for bias detection in higher education learning content. The contributions of this research are fourfold: 1. the creation of multiple novel annotated datasets spanning university news, reading lists, and academic learning resources; 2. the introduction of LLM-based strategies for bias annotation, fine-tuning, and cross-domain evaluation 3. methodological innovations including a structured framework for categorising bias, a hybrid human–AI annotation approach, and a replicable NLP pipeline for demographic and thematic analysis; and 4. the design of a hybrid bias detection system and accompanying web application. This work advances computational approaches to bias detection, provides reproducible resources for future research, and offers practical tools to support greater fairness and equity in higher education online environments.3 0Item Unknown Artificial Intelligence in Routine Non-contrast CT Imaging to Assess Cardiothoracic Structures and Evaluate Clinical Utility(Saudi Digital Library, 2026) Alnasser, Turki; Swift, Andrew; Alabed, SamerBackground: Pulmonary hypertension (PH) is a progressive and life-threatening condition characterised by elevated pulmonary arterial pressure and associated with different diseases, including left heart and lung diseases. Early diagnosis is essential to improve clinical outcomes; however, current diagnostic pathways rely on invasive right heart catheterisation or contrast-enhanced imaging, which are not always feasible in routine clinical practice and are associated with different complications. Non-contrast chest computed tomography (CT) is widely available and frequently performed in patients with PH, many of whom also exhibit coronary artery calcification. However, its full diagnostic and prognostic potentials remain underexplored. Although several manually derived CT measurements have been proposed as imaging predictors, their clinical utility is limited by inter-observer variability and time-consuming analysis. Aim: This thesis investigates the use of artificial intelligence (AI)–driven volumetric measurements of multi-cardiothoracic structures from non-gated, non-contrast CT to enhance both the diagnosis and prognostic assessment of PH and coronary artery calcification. Methods: A multi-cardiothoracic structure and coronary artery calcification AI-based segmentation models were developed at the University of Sheffield using internal and external cohorts from the ASPIRE registry. The models were benchmarked against the gold standard haemodynamic, reference standards, visual assessments, and validated tool (e.g. TotalSegmentator). Results : The developed AI-based models demonstrate high diagnostic accuracy in predicting PH and detecting coronary artery calcifications. The models were comparable to TotalSegmentator and demonstrated higher accuracy than manual clinical practice techniques when evaluated in exploratory testing. AI-derived right atrial volume and coronary artery calcifications are independently associated with increased mortality in PH patients, even after adjustment for age, sex, PH subgroups, and REVEAL score. Conclusion: Automated segmentation and volumetric measurements of multi-cardiothoracic structures in non-contrast CT have the potential to facilitate earlier and accurate diagnosis and prognostic assessment of coronary artery calcification and PH patients, including lung and left heart diseases.18 0Item Unknown Exploring the Utilization of Multimodal LLMs on Personalized Learning in Higher Education(Saudi Digital Library, 2025) Aljumaah, Jana Abdullah; Rana, Muhammad; Rizwan, TaimoorUse of Large Language Models (LLM) is rising in popularity amongst students in higher education. It has transformed the way students tackle their academic work, offering stu- dents powerful new tools for explanation, idea generation, writing support, and independent study. While their adoption has been widely discussed in theory, there remains limited empirical evidence on how students are using these technologies, how they per- ceive their benefits and risks, and how such practices may reshape learning. This dissertation explores the perceptions and practices of 55 UK university students re- garding the use of LLMs in academic work. Using a survey-based methodology, the study examines five key objectives: to identify patterns of LLM use, analyze the purposes for which they are employed, evaluate student attitudes toward their advantages and limita- tions, assess their influence on study habits and independent learning, and provide insights for ethical and effective integration into higher education. The results reveals that students use LLMs as tools for understanding and organizing knowledge such as summarization or concept explanations and clarifications. Secondly, they employ them for supporting their academic writing and production like proofreading, editing, or as a writing assistant. Likert-scale analyses indicate that students generally per- ceive LLMs as both useful and easy to use, consistent with the Technology Acceptance Model (TAM). However, concerns around trust and over-reliance temper this enthusiasm, highlighting a critical awareness of the risks involved. The findings suggest that while LLMs are integrated into study habits and promote efficiency, their long-term value will depend on guidance, ethical frameworks, and digital literacy training. The study contributes to the growing body of research on AI in education by offering evi- dence from the UK universities context. It demonstrates how perceptions of usefulness, ease of use, and reliability shape adoption, and it paves the way for future research, includ- ing cross-cultural comparisons, to better understand the evolving role of LLMs in higher education.40 0Item Unknown Efficient Language Model Compression For Mobile Deployment : A Study on Qwen1.5-1.8b-chat and Phi-3-Mini-4K-Instruct(Saudi Digital Library, 2026) Albalawi, Atheer Mohammed; Aletras, NikosLarge language models (LLMs) offer impressive capabilities but are often too computationally demanding for deployment on resource-constrained devices such as mobile platforms. This project investigates the effectiveness and limitations of compression techniques applied to two representative models, Qwen1.5-1.8B-Chat and Phi-3-Mini-4K-Instruct, using different approaches for each architecture. For Qwen1.5-1.8B-Chat, structured pruning and INT8 quantization were applied, achieving significant size reduction with high benchmark retention (88.4% MMLU, 95.9% ARC-Challenge). However, functional evaluation revealed severe degradation, including deterministic output collapse and loss of generative ability, indicating gaps in benchmark-driven evaluation. In contrast, Phi-3-Mini-4K-Instruct was compressed using EntroLLM mixed quantization and entropy-based methods, which preserved both accuracy and generative behavior, demonstrating greater deployment reliability. These findings highlight that compression outcomes are highly model-dependent and that standard benchmarks may obscure critical failures. This work contributes technical insights by identifying architecture-specific vulnerabilities to compression as well as methodological lessons that underscore the need for comprehensive, deployment-aware evaluation frameworks to ensure reliable LLM performance in practice.10 0Item Restricted Do Generative Chatbot Persuade Users? An Elaboration Likelihood Model Investigation with Cognitive and Emotional Trust(Saudi Digital Library, 2026) Mohanna, Sohaib; AlSurmi, AbdulrahmanGenerative chatbots are increasingly used as consumer information sources, yet their effects on how consumers process and adopt information remain insufficiently understood. It is unclear how the persuasive communication cues used by these chatbots influence a user’s perception of information usefulness and the likelihood of adopting that information. Existing studies in generative chatbots persuasion and information adoption have rarely integrated their theoretical perspectives in this context, focusing mostly on technology acceptance rather than information processing, leaving a gap in understanding how persuasive processes influence in conversations with such systems. Addressing this gap, the study applies an extended Information Adoption Model (IAM), a framework rooted in a dual process theory of persuasion, specifically the Elaboration Likelihood Model (ELM), to generative chatbots dialogue. It investigates how specific persuasive cues in generative chatbots influence users' cognitive and emotional trust, subsequently affecting their perceived information usefulness and adoption decisions. By extending the IAM with a multidimensional view of trust mechanisms (cognitive and emotional), this research contributes to the persuasion process of artificially intelligent human interactions by demonstrating how persuasive mechanisms participate into perceived information usefulness, the critical mediator preceding information adoption in generative chatbots contexts. To investigate these relationships, this research employs a positivist quantitative, theory-testing design. Data from 438 experienced chatbot users were collected, processed for analysis and the proposed model was tested using PLS-SEM. Analysis of the survey data supports the integrated model and reveals that both central and peripheral cues play significant roles via distinct pathways. Central cues, such as high information quality and relevance in chatbots responses, significantly impact cognitive trust and directly enhance consumers’ perceptions of information usefulness. Peripheral cues, including the chatbots human-like characteristics, primarily influence emotional trust. In turn, both cognitive and emotional trust emerge as strong positive predictors of perceived information usefulness and of consumers’ willingness to adopt chatbots information. Interestingly, opposite to the ELM contention of central dominance over peripheral cues, emotional trust exerted stronger influence toward information usefulness. This study makes both theoretical and practical contributions. Theoretically, it demonstrates the value of bridging persuasion and information adoption theories in the field of human–AI communication. By presenting how cognitive and peripheral cues respectively foster cognitive versus emotional trust, the research provides a nuanced understanding of how cognitive and superficial cues through which -provided information becomes persuasive and valuable to consumers. It also extends information adoption models by introducing a multidimensional view of trust perspective and specific chatbots constructs to highlighting that rational credibility and emotional rapport are both critical for consumers’ acceptance of information. In practical terms, the findings offer guidance for designing more effective generative chatbot systems.19 0Item Restricted REGULATING ALGORITHMIC DISCRIMINATION UNDER THE EU AI ACT: EVALUATING BIAS MITIGATION DUTIES FOR HIGH-RISK AND GENERAL-PURPOSE AI SYSTEMS(Saudi Digital Library, 2025) ALSOMALI, ABDULAZIZ; ZIHAO, LIAlgorithmic systems now allocate work, credit, welfare and even police attention (facial recognition systems). They are not ‘neutral’ instruments; they often reproduce and amplify structural disadvantage. This dissertation asks whether the European Union’s Artificial Intelligence Act, when coupled with the Charter of Fundamental Rights and the equality acquis, can prevent and redress such discrimination. This dissertation argues that the Act is normatively necessary but only conditionally sufficient. Its risk architecture, data‑governance duties, documentation and oversight requirements, and the upstream regime for general‑purpose models supply the right legal levers. Constitutional adequacy will materialise only if implementation embeds equality law into technical practice through three cumulative conditions: (i) standards that require context‑specific metric selection justified by proportionality and the availability of less discriminatory alternatives; (ii) supervision with genuine statistical and legal capacity across the system lifecycle; and (iii) remedial pathways that convert logs and technical files into proof under burden‑shifting rules. Thus this paper turns to a functional comparison with the United States and the United Kingdom shows how adverse‑impact doctrine, discovery, and regulator‑led guidance can be harnessed without sacrificing the coherence of the EU model. Followed by Chapter 5 which sets out a concise implementation blueprint and measurable indicators. On that basis, bias mitigation is framed not as ethics, but as a legal duty by which the Act’s success must be judged.15 0Item Restricted Machine Learning for Radiotherapy Treatment of Prostate Cancer(Saudi Digital Library, 2026) Alqarni, Maram; Teresa, Guerrero Urbano; Andrew, KingExternal beam radiotherapy (EBRT) and brachytherapy (BT) are both forms of radiation treatment used for prostate cancer to destroy cancer cells. EBRT applies the radiation externally while BT involves placing radioactive seeds inside the prostate. At Guy’s Cancer Centre, both treatment modalities are performed depending on various factors. Each of the treatment modalities involves different imaging modalities used for treatment planning, delivery and follow-up. However, both have some overlapped clinical tasks such as defining the clinical target volume (CTV) and organs at risk (OARs) from imaging data. The work described in this thesis aims to perform research to promote clinical translation of machine learning (ML) techniques to streamline workflows in EBRT and BT. The first piece of work in this thesis focuses on an ML-based segmentation model for prostate MRI. One of the main challenges affecting clinical adoption of ML in MRI segmentation is the domain shift problem. The findings of this piece of work reveal for the first time the significant impact on model performance of using different acquisition/annotation protocols, even if using the same scanner vendor/field strength. It is shown that training an ML model with data that covers the important sources of domain shift can produce a robust model with good generalisability performance. The next piece of work investigates the possibility of race bias in ML-based prostate MRI segmentation. Through experiments on a controlled dataset of White and Black patients, it is shown that the model performance gap between Black and White subjects is dependent on the level of (im)balance between Black and White subjects in the training data. Again, it is shown that training using demographically balanced data can produce a fair and robust model. The conclusion from both of these pieces of work is that model performance can be robust if the training data is sufficiently diverse, both in terms of image characteristics and patient demographics. Building upon these analyses, the thesis next investigates the clinical utility of a diagnostic prostate MRI model trained on diverse data and externally validates it on in-house clinical data. The evaluation of this model encompasses not only standard quantitative metrics but also measurement of inter-observer variability in manual segmentation and assessments of performance on downstream clinical tasks. Next, the thesis investigates the clinical utility of multi-organ ML-based segmentation models. Here, two models are investigated: one for planning MRI called the “FIMRAa-P” model and another radiotherapy CT model called the “PelvisMA-CT” model. Both models are extensively evaluated quantitatively and qualitatively by five observers. The agreement between the quantitative metrics and the qualitative clinical metrics is also investigated for each clinical structure, revealing generally poor agreement between the two. It is also shown that this agreement is dependent on the structure being segmented and the profession of the clinicians who perform the evaluations. One of the main clinical translation outcomes of this thesis is the deployment of PelvisMA-CT by the Clinical Scientific Computing (CSC) group at GSTFT, and its integration into a contouring application called GSTTAutoSeg. This model is currently being used clinically at Guy’s Cancer Centre and the thesis presents the results of a monitoring and enhancement study based on its ongoing clinical use. Overall, the thesis presents a number of key contributions, all aimed at promoting clinical translation of ML in EBRT and BT. It is hoped that the work performed will accelerate the benefits of ML in radiotherapy treatment planning and delivery and ensure that all patients benefit from the introduction of the thoroughly evaluated new technology.8 0