Postgraduate Theses

Translation of Digitally Enabled Health Initiatives for Human Development

2026-06-16T07:33:45Z

Translation of Digitally Enabled Health Initiatives for Human Development Pujitha Gunawardena, Dinushika Sathsarani Access to digital technologies alone is insufficient to achieve desired improvements in low- and middle-income countries (LMICs). Historically, efforts have focused on exporting solutions from Western settings to LMICs, with limited attention to local social, economic, political, and infrastructural contexts. Such approaches result in failure to achieve expected development outcomes, frustrating local users and posing a high risk for short-lived initiatives. Moreover, they remain inadequate for realising global aspirations for human development as envisioned by the United Nations. This thesis reimagines current approaches to digital development by proposing a translational approach that offers a promising path for the design and local embedding of digitally enabled health initiatives. Chapter 2 provides a conceptual review of human development, leading to a conceptual leap that emphasises translating Information System (IS) artifacts and associated knowledge to better support aspirations for human development. In Chapter 3, a four-phase research design is developed to guide the translation-focused design of digitally enabled health initiatives. Locally anticipated outcomes in Ecuador and Papua New Guinea inform the future design of a virtual care system for their contexts through translation. Chapter 4 investigates how translation facilitates the local embedding of a digitally enabled health initiative in Sub-Saharan Africa. How actors translate IS artifacts and associated knowledge to enhance their relevance within local practices, norms, and cultures is examined. This thesis makes five key contributions: it offers a reimagined approach for digital development research; broadens the understanding of what translation entails in digital development; contextualises and extends the lens of the IS artifact; makes a methodological contribution through a translation-focused research design; and identifies interrelated outcomes of virtual care in remote settings. Includes publication

Capturing Students' Conceptual Change When Exploring Decimals through Dynamic Digital Representations

2026-06-16T02:39:03Z

Capturing Students' Conceptual Change When Exploring Decimals through Dynamic Digital Representations Gorman, Amelia Kate Decimals are of great significance in the primary mathematics curriculum due to their application and use in everyday life. The purpose of this study was to investigate the effectiveness of certain dynamic digital representations in developing students’ knowledge of decimal fractions. Task-based interviews were used with six Year 4 (9-10 years old) students, that incorporated four different dynamic digital representations of decimals. Data collected via video-audio recordings were used to detect shifts in students’ attention while using the digital representations. Attention shifts were analysed using microgenetic methods to determine conceptual changes over time. Findings uncovered specific features of the digital representations that generated productive cognitive con-fusion which prompted changes in students’ understanding of decimal fractions. The unique affordances of each digital tool offered students opportunities to dynamically explore decimal concepts and could be used to enrich the teaching of decimals within the primary mathematics classroom. Includes publication

Confronting Plate Models with the Deep Mantle

2026-06-16T00:59:59Z

Confronting Plate Models with the Deep Mantle New, Thomas Christopher Mantle circulation models, numerical models of whole-mantle convection driven by tectonic reconstructions, are widely used to predict present-day mantle structure and, in turn, to evaluate the reconstructions themselves. Yet many studies compare modelled temperature fields directly with seismic tomography, overlooking nonlinearity in the temperature-velocity relationship, the strongly heterogeneous resolution of seismic imaging, and misfit metrics that saturate when anomalies do not overlap. This thesis develops a robust, physically consistent approach to evaluating competing tectonic reconstructions against tomography. I first develop a transferable comparison methodology, demonstrated with the G-ADOPT finite-element modelling library, comprising: (i) physically consistent conversion of predicted temperatures into seismic velocities; (ii) tomographic resolution operators to filter converted structures prior to comparison; and (iii) the Wasserstein metric to quantify misfit informatively even without anomaly overlap. After benchmarking G-ADOPT models across well-studied subduction regimes, I apply these tools to the collision of the Ontong-Java plateau with the Melanesian arc, where collision timing remains uncertain (~25 vs ~12 Ma). I show that physically consistent post-processing can flip qualitative inferences of slab morphology, shift inferred slab depths by tens of kilometres, change inferred sinking rates by up to 1.2 mm/yr, and laterally displace slab anomalies by up to 639 km. Benchmarking shows G-ADOPT reproduces mantle structure well where slab rollback dominates, but predicts incorrect dip polarity where trenches advance. The case study strongly supports the older collision scenario: only the ~25 Ma model reproduces slab material beneath the Melanesian arc, with 10-28% lower misfit, supporting a role for the collision in regional plate reorganisation. These results highlight the value of publishing tomography models alongside resolution operators. Includes publication

AI-driven Defect Detection and Multi-Agent Disaster Response for Civil Infrastructure

2026-06-15T23:41:17Z

AI-driven Defect Detection and Multi-Agent Disaster Response for Civil Infrastructure Chen, Zhaohui Rapid and reliable assessment of civil infrastructure is essential for effective disaster response. Traditional inspection relies on manual, expert-driven surveys, which are time-consuming, resource-intensive, and difficult to scale under limited accessibility and evolving hazards. Although computer vision enables automated damage detection, most approaches focus on isolated perception tasks and lack support for system-level reasoning. This thesis develops an AI-driven framework that integrates robust local-scale perception with structured global-scale reasoning. At the local scale, efficient vision models are designed for defect detection under realistic conditions. A Transformer-based crack segmentation model with average pooling improves robustness in noisy environments, while a Robust Feature Knowledge Distillation (RFKD) framework transfers noise-resilient representations from teacher to lightweight student models for deployment. Vision Mamba architectures are further explored to improve scalability for high-resolution inspection. At the global scale, a multi-agent framework transforms perceptual outputs into actionable insights through role-based task decomposition, information integration, and collaborative reasoning, enabling coherent situational awareness with human-in-the-loop support. Experiments demonstrate improved robustness and scalability over conventional methods. Challenges remain in handling uncertainty and ensuring reliable long-horizon reasoning. Future work will focus on uncertainty-aware models and tighter integration with sensing and decision-support systems. Includes publication

Investigating the Mechanisms and Applications of Liquid Metal-Biomolecule Interactions

2026-06-15T23:25:43Z

Investigating the Mechanisms and Applications of Liquid Metal-Biomolecule Interactions Liu, Li Gallium (Ga)-based liquid metals have attracted increasing attention due to their metallic conductivity, fluidity, and unique surface chemistry. However, their interactions with biological macromolecules remain insufficiently understood. This thesis investigates the interfacial interactions between liquid Ga droplets and biomolecules, focusing on protein self-assembly and nucleic acid reactivity. First, a self-standing soy protein isolate (SPI)-Ga composite film is developed to study interactions between Ga droplets and protein fibrils. Protein fibrils reduce Ga surface oxidation, promote droplet coalescence, and facilitate conductive pathway formation without disrupting the β-sheet-rich fibrillar network. The resulting composites exhibit combined electrical conductivity and mechanical robustness, enabling applications in gas sensing and electrically stimulated antibacterial activity. This thesis then investigates the interaction between Ga droplets and DNA. The results demonstrate that Ga droplets can cleave DNA phosphodiester bonds, with preference for adenine- and thymine-rich sequences. Mechanistic studies show that the activity originates from interfacial electron transfer associated with Ga surface oxidation, generating reactive species that cleave the DNA backbone while largely preserving nucleobase integrity. The nuclease-mimicking activity can also be tuned through droplet characteristics and external stimuli. Together, this thesis establishes a fundamental understanding of how liquid Ga interfaces interact with protein fibrils and nucleic acids, providing guidance for the design of liquid metal-based biohybrid materials and future biomedical applications. Includes publication

Optimizing Large Language Models: Algorithmic Advancements and Model Design Strategies

2026-06-15T23:06:56Z

Optimizing Large Language Models: Algorithmic Advancements and Model Design Strategies Bie, Fengxiang Large Language models have achieved remarkable performance across diverse tasks, but face two critical deployment challenges: (1) the key-value (KV) cache memory bottleneck that limits model deployment in resource-constrained environments, and (2) the sequential autoregressive generation latency that reduces inference throughput and user experience. This thesis presents two complementary contributions addressing these distinct challenges. First, CARE (Covariance-Aware and Rank-Enhanced) tackles the KV-cache memory bottleneck by converting pretrained Grouped Query Attention (GQA) models into memory-efficient Multi-Head Latent Attention (MLA) architectures. Unlike naive SVD approaches that ignore activation patterns, CARE introduces activation-preserving factorization using covariance-weighted SVD and adaptive rank allocation via water-filling algorithms. Second, Infinigram-based speculative decoding addresses inference latency by leveraging large-scale n-gram statistics to predict multiple tokens in parallel, achieving significant speedup through CPU-optimized data structures and confidence-based acceptance strategies. Experimental results on Llama-3.1-8B demonstrate that CARE achieves up to 331% relative improvement in zero-shot accuracy over baseline conversion methods while maintaining identical KV-cache footprint. Post-conversion healing fully recovers original model performance with minimal fine-tuning. Infinigram delivers significant inference speedups across various sequence lengths and batch sizes, with acceptance rates improving for longer context matches and higher-frequency patterns. This work contributes novel methodologies combining model design strategies and algorithmic advancements for efficient large generative model deployment, providing practical solutions to key memory and computational challenges without compromising model capabilities.

Fast Algorithms for Fréchet-Based Similarity

2026-06-15T22:50:08Z

Fast Algorithms for Fréchet-Based Similarity Huang, Zijin Polygonal curves arise naturally when recording movement data from GPS traces, animal tracking, or eye-gaze measurements. A fundamental task is to compare such curves using a formal notion of similarity. The Fréchet distance is a widely used metric for this purpose, as it respects both the geometric proximity and the traversal order of two curves. This thesis develops faster algorithms for four Fréchet-based similarity problems, each using the freespace diagram as its central tool. In Chapter 2, we study the Fréchet distance under geometric transformations, which is essential when comparing shapes regardless of their position or orientation. We present the first improvement to the decision problem for a wide class of transformations, including translations, rotations, and scalings. In Chapter 3, we consider the Fréchet edit distance, which allows inserting or deleting vertices before measuring the continuous Fréchet distance. This variant is motivated by the need for robustness against outliers and noise in real-world trajectory data. We provide faster algorithms for several edit models, improving on prior bounds by up to a factor of k · n. In Chapter 4, we study subtrajectory clustering on c-packed trajectories, a realistic model of movement data with bounded self-overlap. We present a near-linear time approximation algorithm for the subtrajectory cluster problem, circumventing the conditional cubic lower bound for general curves. In Chapter 5, we connect the partial weak Fréchet similarity to shortest paths in weighted planar regions. We construct a near-linear size approximate spanner for the 0/1/∞ weighted region problem, and apply it to obtain an approximation algorithm for the partial weak Fréchet similarity.

BayeSpace - Applying Bayesian Statistics to Environmental and Ecological Phenomena

2026-06-15T22:33:28Z

BayeSpace - Applying Bayesian Statistics to Environmental and Ecological Phenomena Davis, Samuel Caradog Environmental and ecological phenomena are inherently complex, dynamic, and uncertain — requiring analytical tools that can generate accurate predictions while transparently quantifying uncertainty. This thesis introduces BayeSpace, a modular Python framework for applying Bayesian inference to spatiotemporal environmental and ecological modelling. It addresses key limitations in existing approaches, including fragmented workflows, steep learning curves, and model visualisation and comparison. BayeSpace integrates two core Bayesian methodologies: Bayesian Regression (BR) for problems with known functional forms, and Gaussian Process Regression (GPR) for flexible, non-parametric modelling. Built on libraries such as NumPyro and scikit-learn, it provides core classes for model definition, prior and likelihood specification, sampling, and visualisation. Key innovations include automated experiment tracking, domain generation, data simulation, kernel and transformation flexibility, and robust visualisation tools. The framework's capabilities are demonstrated through several applications. In marine cloud brightening research, BR was used to infer atmospheric dispersion parameters from real-world plume data over the Great Barrier Reef, suggesting oceanic dispersion may be more intense than terrestrial models predict. This case study illustrates BayeSpace's ability to explore complex phenomena, identify model limitations, and understand asymptotic behaviour. In species distribution modelling, GPR effectively handled spatiotemporal occurrence data, with model performance strongly correlating with total grid coverage. BayeSpace's iteration ability and flexibility were vital in designing a one-size-fits-all framework for species distribution modelling. Validation using simulated and real-world datasets confirmed BayeSpace's accuracy in linear cases while revealing challenges in highly non-linear regimes where complex posteriors affect sampler convergence.

Essays on Machine Learning in Economics

2026-06-15T10:48:59Z

Essays on Machine Learning in Economics Deng, Sinan This thesis presents three studies exploring how machine learning can be applied to understand and model complex phenomena in economics. Chapter 2 investigates how machine learning can be used to examine geographic diversity within global economics research. Chapter 3, published in Energy Economics (2024), develops a seasonal deep learning model for Great Britain’s electricity imbalance prices, showing that incorporating seasonality improves forecasting accuracy. Chapter 4 extends this framework by applying distributional deep learning methods to model full price distributions and distinguish between aleatoric and epistemic uncertainty. Includes publication

Engineering entanglement in trapped ion quantum harmonic oscillators using spin-dependent interactions

2026-06-15T10:33:09Z

Engineering entanglement in trapped ion quantum harmonic oscillators using spin-dependent interactions Millican, Maverick James Quantum harmonic oscillators are a central resource in quantum technologies. Coupling this resource to a two-level system gives rise to spin-oscillator dynamics that can be used for system calibrations, enhanced sensitivity to signals, and the preparation of oscillator entangled states. Chapter 4 introduces a motional-frequency calibration protocol based on a time-reversal, state-dependent force sequence that maps the effects of motional-frequency miscalibration onto a narrow spin-response feature. The chapter also discusses how thermally occupied motional states enhance sensitivity to particular Hamiltonian terms, making them effective probe states. A two-point feedback servo with processing performed on a field-programmable gate array, or FPGA, is used to track the interaction resonance in real time, enabling sub-second updates and continuous correction of frequency drift during data acquisition. Chapter 5 realizes phase-insensitive displacement sensing by embedding spin-dependent squeezing in a time-reversal interferometry sequence. The method prepares a spin-correlated superposition of orthogonally squeezed motional states, yielding a spin-observable response used to estimate displacement amplitudes while remaining agnostic to the signal phase. Experiments with vacuum and number-state probes demonstrate enhanced sensitivity with increasing motional energy. Chapter 6 extends the spin-oscillator toolkit to multi-oscillator control. Optimized phase modulations applied to Jaynes-Cummings and anti-Jaynes-Cummings couplings between a single spin and two motional modes synthesize an effective two-mode squeezing interaction, enabling preparation of two-mode squeezed vacuum states and a non-Gaussian superposition of those states. Joint phase-space tomography reveals multi-mode correlations, and continuous-variable entanglement is certified by the Einstein-Podolsky-Rosen criterion and a Clauser-Horne-Shimony-Holt-type continuous-variable Bell test. Includes publication

Mesenchymal stem cell senescence: Mechanisms and rejuvenation strategies

2026-06-15T07:40:23Z

Mesenchymal stem cell senescence: Mechanisms and rejuvenation strategies Zhang, Yiran Cellular senescence is a key contributor to organismal ageing and reduces the regenerative capacity of mesenchymal stem cells (MSCs), limiting their therapeutic potential in treating age-related diseases. This thesis aimed to investigate the mechanisms and functional consequences of MSC senescence and to develop strategies to rejuvenate senescent cells. Using adipose-derived mesenchymal stem cells (ASCs), this study first examined the effects of the chemotherapeutic drug doxorubicin (DOX) on senescence. Both short- and long-term DOX exposure induced senescence; however, while long-term exposure impaired osteogenic differentiation, short-term exposure enhanced osteogenesis. Transcriptomic analysis identified insulin-like growth factor 2 (IGF2) as a potential regulator of this dual effect, and IGF2 treatment restored osteogenic potential in senescent ASCs, suggesting a beneficial role of transient stress-induced senescence. To explore rejuvenation strategies, this thesis investigated both biomaterial and drug delivery approaches. Osteopontin (OPN), a major extracellular matrix protein, was shown to reverse senescence-associated phenotypes in replicative senescent ASCs, restoring proliferation, osteogenic differentiation, trophic support, and regulating cell morphology. In parallel, a liposome-based drug delivery system was developed to encapsulate nicotinamide mononucleotide (NMN), an anti-senescence drug. Optimised formulations achieved high encapsulation efficiency and reduced the expression of senescence markers in vitro. Together, these findings advance understanding of MSC senescence and identify OPN-based biomaterials and liposomal delivery systems as promising strategies for MSC rejuvenation and regenerative medicine.

Development of a sustainable acid resistant mortar for back-filling of impressed current cathodic protection (ICCP) anodes

2026-06-15T05:55:51Z

Development of a sustainable acid resistant mortar for back-filling of impressed current cathodic protection (ICCP) anodes Fatemi Nayeri, Sayed Hamid Reza Reinforced concrete (RC) structures in marine environments are highly susceptible to chloride-induced corrosion, for which impressed current cathodic protection (ICCP) is widely applied. The performance of ICCP systems is governed by the durability and electrochemical behaviour of the anode back-fill mortar. Field observations from New South Wales, Australia, show that conventional Portland cement-based mortars can degrade due to electrochemically induced acidification at the anode interface, highlighting the need for more stable and compatible materials. This thesis investigates ICCP-induced acidification through field sampling, laboratory characterisation, and electrochemical testing. Analyses of mortars retrieved from operational marine bridges show that anodic reactions generate acidic species that dissolve calcium-rich phases, promote gypsum formation, and degrade the binder, resulting in non-uniform deterioration. These mechanisms were reproduced under accelerated laboratory conditions. Based on this understanding, hybrid geopolymer mortars incorporating supplementary cementitious materials (SCMs) were developed as acid resistant alternatives. Laboratory results show improved resistance to acidic exposure, reduced mass loss, and enhanced microstructural stability. Field implementation confirmed material integrity and compatibility with electrochemical requirements under marine exposure. Transport properties were also assessed. While higher SCM content improved durability, it increased resistivity, highlighting a trade-off between durability and electrochemical performance. Ionic conductivity governed polarisation behaviour, and its optimisation improved performance in atmospheric zones, while conductive additives were unnecessary in submerged conditions. Overall, this research establishes an exposure-specific framework for designing ICCP back-fill mortars that balance durability and electrochemical functionality in marine environments.

Carbon and Nutrient Interactions in Cereal-Legume Intercropping Systems: Impacts of Phosphorus Fertilization

2026-06-15T05:31:17Z

Carbon and Nutrient Interactions in Cereal-Legume Intercropping Systems: Impacts of Phosphorus Fertilization Rahman, Md Zillur Phosphorus (P) deficiency is a major constraint to agricultural productivity in low-input systems where nutrient availability limits crop growth and resource-use efficiency. Although cereal-legume intercropping is widely promoted to improve nutrient acquisition and productivity, the mechanisms by which P availability regulates nutrient uptake, biological nitrogen fixation (BNF), nitrogen competition, and belowground carbon (C) allocation remain poorly understood. This thesis investigated the effects of P fertilization on productivity, nutrient acquisition, and carbon-nitrogen interactions in cereal-legume systems through a global meta-analysis and isotope-based experiments in a P-limited soil. A meta-analysis showed that P fertilization significantly increased crop yield and N and P uptake in both monocropping and intercropping, while also improving land-use efficiency in intercrops. Experimental studies using wheat-chickpea intercrops demonstrated that P availability strongly regulated belowground C allocation and BNF. Phosphorus fertilization increased chickpea biomass, P uptake, and BNF, while reducing root C allocation, indicating a shift towards symbiotic N acquisition. In contrast, wheat maintained root C investment regardless of cropping system. A 15N-labelling experiment revealed that wheat was more competitive for nitrate uptake than chickpea, particularly under P fertilization, whereas ammonium acquisition was similar between species. Fine-root traits were important predictors of nutrient uptake under P limitation. A 13CO2 pulse-chase study showed contrasting P-acquisition strategies, with chickpea maintaining high P uptake with minimal changes in rhizodeposition, while wheat relied on increased root C allocation and rhizodeposition. These findings provide new insights into nutrient acquisition and belowground C dynamics in cereal-legume intercropping and highlight opportunities to improve P-use efficiency in sustainable low-input cropping systems. Includes publication

Causes, Mechanisms, and Mitigation of Socially-Induced Nocebo Effects

2026-06-15T04:31:53Z

Causes, Mechanisms, and Mitigation of Socially-Induced Nocebo Effects Saunders, Cosette Emilie The nocebo effect refers to adverse symptoms that arise in response to treatment but cannot be explained by its active properties. Despite their major clinical, social, and economic burden, nocebo effects remain understudied. Social learning, where people learn by observing or interacting with others, is a potent pathway for nocebo effects. In healthcare, where treatment experiences can be shared face-to-face and through mainstream and social media, understanding socially-acquired nocebo effects matters. This thesis examined the causes, mechanisms, spread, and reduction of socially-induced nocebo effects. Chapter 1 introduces nocebo effects, induction pathways, and key mechanisms. Chapter 2 presents a review and meta-analysis showing that observing adverse symptoms in another person produces medium-to-large nocebo effects compared with no treatment, with effects similar to classical conditioning and larger than explicit instruction. Prior studies, however, only tested cases where model and observer had identical experiences. Chapters 3 and 4 tested whether nocebo effects arise when the model’s experience differs from the observer’s. Using a virtual reality model of cybersickness and a simulated clinical paradigm, these studies showed that social nocebo effects generalise to similar treatments and contexts. This suggests social expectations are not limited to the modelled intervention, widening the scope of harm. Chapters 3 and 5 tested ways to reduce social nocebo effects. Choice was not found to reduce socially elicited cybersickness. In contrast, positive social modelling, where side effect warnings were paired with a peer reporting no side effects, reduced symptom severity. This gives initial evidence that social learning can also counter nocebo formation. Overall, this thesis advances understanding of socially-induced nocebo effects and shows that continued work is needed to develop and test ways to reduce their burden on individuals and the wider community.

Leveraging deep learning to enable real-time beam-view image-guided radiotherapy

2026-06-15T03:51:29Z

Leveraging deep learning to enable real-time beam-view image-guided radiotherapy Chrystall, Danielle Maria Real-time image-guided radiotherapy (IGRT) is essential for minimising the clinical impact of patient motion during radiotherapy. Existing IGRT systems often rely on additional imaging dose or specialised, expensive technology to enable continuous intrafraction monitoring. Beam-view imaging offers a promising alternative, enabling real-time tumour monitoring directly in the treatment beam without additional imaging dose and with broad compatibility on standard linear accelerators (linacs). However, its clinical use is limited by treatment-beam occlusions and poor megavoltage (MV) image quality. This thesis aims to develop, implement, and evaluate a deep learning-enabled beam-view IGRT framework that is safe, accurate, and compatible with standard linacs. Novel deep learning approaches were leveraged to overcome the aforementioned challenges facing beam-view imaging. Three research objectives are addressed: (i) develop deep learning-enabled beam-view target tracking approaches for abdominopelvic and thoracic treatment sites; (ii) experimentally evaluate real-time beam-view marker tracking using an anthropomorphic pelvic phantom, and develop associated workflows and patient-specific quality assurance procedures to facilitate safe clinical deployment for prostate cancer radiotherapy; and (iii) clinically implement and evaluate real-time beam-view marker tracking during prostate stereotactic body radiotherapy (SBRT). Real-time beam-view IGRT has been developed and investigated, with clinical feasibility demonstrated for prostate SBRT. Key implementation barriers are addressed, establishing a foundation for broader clinical adoption of real-time beam-view IGRT on standard linacs.

Multimodal Emotion Elicitation and Recognition in Virtual Reality

2026-06-15T03:14:16Z

Multimodal Emotion Elicitation and Recognition in Virtual Reality Kuang, Zheyuan Virtual Reality (VR) has been effectively used for eliciting emotions, yet most research focuses on the intensity of affective responses rather than on how interaction influences those experiences. To address this gap, this thesis advances a validated VR emotion-elicitation dataset through two extensions. First, we add a new high-arousal, high-valence scene and validate its effectiveness in a within-subject study (N=24). Second, we create interactive and non-interactive versions of each scene to examine the impact of interaction on emotional responses. We evaluate interaction using subjective ratings and physiological signals. Our evaluation study (N=84) shows that interaction not only amplifies emotions but also modulates them in context, supporting coping in negative scenes and enhancing enjoyment in positive scenes. Multimodal Emotion Recognition (MER) increasingly depends on fine-grained, evidence-grounded annotations, yet inspection and label construction are hard to scale when cues are dynamic and misaligned across modalities. This thesis presents an LLM-assisted toolkit that supports multimodal emotion data annotation through an inspectable, event-centered workflow. The toolkit aligns heterogeneous recordings, visualizes modalities on a shared timeline, and packages synchronized keyframes and time windows as traceable event packets. It then uses modality-specific tools and prompt templates to draft structured annotations for analyst verification and editing. Building on the dataset extensions and annotation, this thesis further investigates MER modeling approaches in VR that integrate behavioural and physiological signals from VR headsets and wearable sensors. We introduce an LLM-based Mixture-of-Experts (MoE) framework, where experts specialize in different modalities and a router assigns weights to experts for each event. The goal is to connect predictions to traceable multimodal evidence and support interpretation of affective cues in interactive VR.

Learning Theory for Transformers: An Operator-Learning Viewpoint

2026-06-15T02:59:27Z

Learning Theory for Transformers: An Operator-Learning Viewpoint Liu, Peilin Large language models (LLMs) have reshaped the foundations of artificial intelligence research and the modes of interaction between human cognition and machine intelligence. Their influence extends further still, transforming the scientific tools through which we interrogate and model the physical world. Underlying most of these achievements and breakthroughs is a dominant architecture: the Transformer. Although the Transformer was proposed nearly a decade ago, established mathematical frameworks remain insufficient to explain the complex phenomena observed in practice with Transformer-based networks, particularly large language models. This thesis offers a principled theoretical foundation for understanding the remarkable capabilities these models exhibit, grounded in a central argument that the Transformer performs operator learning during pretraining over vast text corpora. Our analysis reveals the nature of pretraining and in-context learning mechanisms of efficient Transformer structures in an operator learning framework. Transformers maps each context distribution to a response function for queries and with more samples from context distribution, they can recover information as much as possible to get a better response function with fixed and pretrained weights without any update.

If 'I am woman', what is man? British masculinity and the Women's Liberation Movement, 1970-1990

2026-06-12T02:40:33Z

If 'I am woman', what is man? British masculinity and the Women's Liberation Movement, 1970-1990 Wallhead, Emma Jane This thesis examines the theory and analysis by the Women’s Liberation Movement (WLM) about men - including their sons, husbands, fathers, friends and others - in the context of the project of liberation. The Women’s Liberation Movement (WLM) from the late 1960s has been widely recognised as naming men as oppressors in the systemic subordination of women. Beyond this, however, there has been little attention paid to the specific critique of men and masculinity that was made by this influential movement. Filling this gap, this thesis uncovers a consistent thread of thinking in the critique of men and masculinity that was sustained across campaigns and across the duration of the WLM. That thinking understood masculinity as a mechanism to assert power over others but, more explicitly, pointed to specific traits such as selfishness, hierarchical thinking, aggression and detachment as socially endorsed norms of masculinity. Rather than creating new forms of ‘masculinity’, feminists argued that, if systems of oppression were to be dismantled, gender categories such as masculinity would need to be abandoned in favour of a more egalitarian world in which women and men might coexist with a shared commitment to collective responsibility, regard for others, nonviolence and care.

ESSAYS ON Strategic Control and Consumer Response: Insights from Vertical Restraints and Promotional Tools

2026-06-11T03:15:10Z

ESSAYS ON Strategic Control and Consumer Response: Insights from Vertical Restraints and Promotional Tools Chen, Chia-Ying This thesis examines how location shapes market outcomes through its influence on both consumer behaviour and distribution channel design. The first essay investigates the effectiveness of electronic coupons using transaction, review, and mobile GPS data. The results show that consumers located farther from a focal restaurant are more likely to leave reviews after coupon redemption. E-coupons also generate positive spillover effects by increasing subsequent review activity at nearby restaurants. These effects vary across mobility segments, with transit-oriented consumers exhibiting stronger exploratory behaviour than consumers associated with workplaces or educational institutions. The second essay examines the interaction between exclusive dealing and exclusive territories in automobile distribution channels. Using multi-level data from the Chinese automobile industry, the study finds that the two contractual arrangements operate as complementary governance mechanisms in luxury markets. Dealer size negatively affects the likelihood of exclusive dealing, particularly for luxury brands. Overall, this thesis demonstrates how spatial considerations influence firm performance, market coordination, and competitive outcomes through both consumer mobility patterns and geographic distribution strategies.

Responsible Management of AI Use in Organizations: Case Studies of Strategic AI Capabilities, Risk Controls, and Knowledge Management

2026-06-10T01:22:43Z

Responsible Management of AI Use in Organizations: Case Studies of Strategic AI Capabilities, Risk Controls, and Knowledge Management Wang, Yichen Organizations are increasingly investing in artificial intelligence (AI), yet many still struggle to translate its technical potential into reliable, and strategically aligned organizational outcomes. These challenges have intensified as AI evolves from static predictive analytics to dynamic and opaque systems embedded in core organizational processes. Existing responsible AI approaches mainly provide high-level ethical principles. As such, an important research question remains: How can organizations achieve responsible management of AI use? To address this question, this thesis presents three qualitative case studies across different AI technologies and organizational contexts, contributing to explain how organizations can develop capabilities, governance mechanisms, and control approaches for responsible use. In particular, Study 1 examines how a social-media platform develops strategic AI capability when implementing predictive AI systems. The findings identify a cyclical capability development process that enables organizations to cultivate informed agility, controlled efficiency, and anticipatory resilience in response to evolving predictive AI outputs. Study 2 investigates dynamic learning recommendation systems across three leading social-media platforms and develops a cybernetic control approach for governing black-box AI systems through buffering, feedforward, and feedback controls. This study demonstrates how organizations maintain model performance and reliable control under conditions of continuous model evolution and learning instability. Study 3 examines a cybersecurity organization adopting generative AI for organizational knowledge management. Drawing on the SECI model of knowledge creation, this study proposes a phase-dependent AI alignment approach and identifies alignment mechanisms that support accountable, and domain aligned AI-generated knowledge outcomes.

Estimating Output Gaps in Open Economies

2026-06-09T07:51:34Z

Estimating Output Gaps in Open Economies De Gorostiza, Gilliane Angela Estimating the output gap remains a critical challenge for macroeconomic policy due to data limits, reporting lags, and global shocks. This dissertation extends the Beveridge-Nelson (BN) decomposition framework across three chapters to provide more reliable, informative, and timely indicators of economic slack for emerging Asian economies and Australia, demonstrating that multivariate and mixed-frequency BN frameworks improve real-time policy decision-making. The first chapter shows that the BN filter provides more reliable estimates for emerging Asian economies than Hodrick-Prescott, Christiano-Fitzgerald, or Hamilton filters. Cyclical consumption is more volatile than the output gap, and less than one-third of GDP growth fluctuations stem from trend shocks, countering the "cycle is the trend" view. Crucially, BN estimates suffer from smaller, less frequent revisions during major economic shifts. The second chapter reveals that while traditional domestic slack measures are uninformative for Southeast Asian economies due to structural issues and informal employment, financial and external variables are highly relevant. Financial factors dominated during the Asian and Global Financial Crises, with external variables often explaining a larger share of cyclical fluctuations than domestic output. The third chapter applies a mixed-frequency framework to Australia. It finds that the labor market's intensive margin (aggregate hours worked) offers a more significant informational contribution than headline unemployment. Furthermore, the Trade Weighted Index (TWI) provides informational value nearly equivalent to the entire financial or macroeconomic sectors combined. While domestic shocks drive most Australian fluctuations, the Global Financial Crisis was largely driven by foreign shocks. Finally, a weekly TWI indicator allows for more timely updates but does not improve nowcast accuracy relative to a monthly frequency.

Writing a Migrant Body: Identity Predicament and Resistance in Sinophone Fiction by Chinese Migrant Women in Australia 1996-2004

2026-06-09T07:13:54Z

Writing a Migrant Body: Identity Predicament and Resistance in Sinophone Fiction by Chinese Migrant Women in Australia 1996-2004 Ye, Su This thesis examines the migrant identities of mainland Chinese women in 1990s Australia through Sinophone fictional works by migrant women writers, centring the migrant body as a core site of gendered, racialized, and transnational power negotiation, exploitation, and resistance. Spanning five interwoven chapters, the study traces a thematic progression from male migrants’ socio-economic precarity and bodily marginalization to female migrants’ layered quests for belonging, sexual subjectivity, marital agency, and cross-cultural solidarity. Chapter One establishes an analytical baseline via metamorphosis and gaze subversion, highlighting the vulnerable male migrant body in women’s writing. Chapter Two explores “return” as a response to rootlessness, revealing the psychological un-recoverability of the homeland and an alternative return to Eastern Buddhist-Taoist philosophy. Chapter Three examines female sexual subjectivity as emerging not through linear Western emancipation but via imitation, sacrifice, rupture, and utilitarian coping. Chapter Four analyses migrant marriages as sites of power imbalance, where the female body becomes a locus of gendered labour displacement, exploitation, and resistance. Chapter Five explores “sisterhood” from internalized misogyny to cross-racial, cross-class solidarity. The thesis reveals that the conflicting roles of the literary “body” align with the complexity of Chinese women’s migrant identities, and that “writing a migrant body” is a unique tool to expose and resist multiple pressures. It argues that Sinophone fiction by 1990s migrant women reclaims the female body from Western Orientalist stereotypes and redefines migrant identity through nuanced portrayals of resistance and agency, revealing female solidarity as a vital form of empowerment. Migrant identity and feminist consciousness are iterative processes, forged through repeated engagement with marginalization across all spheres of migrant life.

Turning the gaze towards the monstrous: Alternative visions of humanity in the works of Virginie Despentes, Julia Ducournau and Paul B. Preciado

2026-06-09T06:02:26Z

Turning the gaze towards the monstrous: Alternative visions of humanity in the works of Virginie Despentes, Julia Ducournau and Paul B. Preciado Smith-Davies, Beaudicea This thesis examines the work of three contemporary Francophone artists: Virginie Despentes, Julia Ducournau, and Paul B. Preciado. It uses the monster of Mary Shelley’s Frankenstein as a model to analyse the monsters in the texts and films of these three artists. It argues for an all encompassing and universal monstrosity, which transcends binary oppositions and speaks to the whole of humanity.

The Unrealised Potential of AI Solutions for Pasture-based Dairy Systems

2026-06-09T02:58:19Z

The Unrealised Potential of AI Solutions for Pasture-based Dairy Systems Azubuike, Blessing Nnenna Efficient management of pasture-based dairy systems can benefit substantially from integrating individual cow supplementation, real-time pasture monitoring, and grazing event detection, but traditional herd-level approaches cannot address the biological variation and dynamic pasture growth inherent in commercial operations. This thesis developed and validated AI and machine learning methods across three interconnected domains, establishing empirical foundations for integrated precision management. Feeding optimisation was addressed through two studies. A Random Forest model combined with Differential Evolution reallocated concentrate to 81 cows across 91 days, achieving an 8% theoretical milk yield increase without additional feed cost (Chapter 3). Four evolutionary algorithms applied to 1,053 training cows and 165 optimised cows over 30 days achieved 6.63 to 8.64% theoretical yield improvements, with NSGA-II outperforming all others across 10 runs (Chapter 4). These gains are predictive estimates; controlled field validation remains necessary. Pasture biomass estimation was addressed through satellite and smartphone-based approaches, both validated against rising plate meter ground truth. An XGBoost model trained on Sentinel-2 imagery from 16 farms achieved R² = 0.70 and MAE = 216 kg DM/ha, outperforming the commercial Pasture.io platform (Chapter 5). These estimates supported automated grazing event detection across 12 farms, where Random Forest achieved within-year F1 = 0.878 and One-Class Support Vector Machine achieved cross-year F1 = 0.692, outperforming supervised models by 7.6% on Year 2 data despite average 24.2% supervised degradation (Chapter 6). Smartphone imagery achieved R² = 0.561 as a low-cost complement sensitive across high-biomass ranges where satellite indices saturate (Chapter 7). Each domain functions independently on commercial farms, though the integrated system linking all three in real time remains the frontier for future research.

Linguistic Landscapes, Assemblages, and Affective Regimes in Chongqing’s Public Transport Hubs: From Transit Spaces to Meaningful Places

2026-06-08T23:02:35Z

Linguistic Landscapes, Assemblages, and Affective Regimes in Chongqing’s Public Transport Hubs: From Transit Spaces to Meaningful Places Liao, Ke This thesis examines how the linguistic landscape (LL) shapes the social functions of public transport hubs and generates patterned affective experiences in Chongqing, a megacity in Southwestern China. Responding to the largely quantitative focus of prior LL research in China, this thesis advances an interpretive and theoretical informed analysis of how signage mediates relations among people, space, affect, and mobility. Drawing on three rounds of large-scale data collection, this thesis first maps the categories, spatial distribution, and linguistic composition of signage across six major transport modes. This quantitative overview identifies key semiotic features and notes changes in the institutional and social functions of these transport hubs. Building on this foundation, an in-depth qualitative analysis of Chongqing North High-Speed Railway Station employs Pennycook’s (2017) assemblage and Scollon and Scollon’s (2003) geosemiotics framework to examine the dynamic interactions among linguistic and semiotic resources, passengers, and differentiated spatial zones. The analysis is further extended to Jiangbei International Airport, where Wee and Goh’s (2019) concept of affective regimes is integrated with Bourdieusian notions of affect and capital to elucidate how passengers’ emotions are structured, circulated, and rendered socially productive across interconnected online and offline contexts. Overall, this thesis demonstrates how LL transforms transport hubs from sites of transit into multifunctional and meaningful places through co-constitutive sign–people–space relations. It also shows how affect is institutionally organised and implicated in the production of social functions within regimes of mobility. Empirically, this thesis contributes a rich and systematic corpus to LL research on China and transport infrastructures; theoretically, it advances the integration of assemblage and affect in LL scholarship.

Towards a Framework for Community Stakeholder Engagement with Infrastructure Projects Through Social Media

2026-06-08T11:24:09Z

Towards a Framework for Community Stakeholder Engagement with Infrastructure Projects Through Social Media Zhang, Jingbo In infrastructure projects, systematic engagement with the perspectives and positions of community stakeholders is usually challenging. In this study, social media is introduced as a platform for observing and gathering the voices and attitudes of community stakeholders. Based on traditional stakeholder engagement theory, this research employs naturalistic inquiry to collect data on the organic interactions between stakeholders and projects on social media. It analyses and summarises stakeholder engagement patterns on social media, introducing a novel framework of stakeholder sentiment and emotion analysis. This theoretical framework includes a new social media dialogue model. It employs two main analytical tools: categorical grouping of online community stakeholders and a stakeholder emotion matrix based on social media data. The model also categorises the online response strategies for projects based on the output of these tools. This study supplements current stakeholder engagement theory by providing a framework to guide online stakeholder engagement. The theoretical framework outlines the general environment for online stakeholder engagement, offers essential elements and steps for project stakeholder dialogues, and provides appropriate theories and methods for different dialogue stages and steps. It offers new perspectives and theories for future research on online stakeholder engagement. The novel social media stakeholder framework proposed in this study can be applied to project stakeholder engagement practices, providing new analytical tools and response strategies for studying community stakeholder engagement via social media, which will enable practitioners to use social media to carry out community stakeholder engagement for projects more efficiently.

Young active galaxies across the radio spectrum

2026-06-10T06:29:01Z

Young active galaxies across the radio spectrum Kerrison, Emily Florence The lifecycle of active galaxies is an open question in modern astronomy. In particular, not all active galaxies are active radio galaxies, in possession of synchrotron-emitting jets, and it is not obvious why. This thesis focuses on these active radio galaxies, and was completed as part of the First Large Absorption Survey in HI conducted with the Australian SKA Pathfinder (ASKAP-FLASH). Early on, we realised many HI detections in ASKAP-FLASH were made towards so-called 'Peaked Spectrum' sources, active galaxies in which the synchrotron jets are still embedded in the dense, nuclear gas of their host galaxy. The thesis presents a new Bayesian framework for identifying peaked spectrum sources using pre-existing radio survey data, RadioSED. Applying RadioSED to a test field, we increase the number of known peaked spectrum sources by more than an order of magnitude in that area through careful treatment of pre-existing datasets. We investigate the multiwavelength properties of this sample, and identify that many of them are distant, making them interesting probes of the physical conditions at cosmic noon and beyond. We then take early results from the ASKAP-FLASH Pilot Surveys and use a new pipeline for mock observations, SANGRiA, to determine whether we can recover the high observed HI detection rate towards compact jets in simulations, under the assumption that it is a purely geometrical effect. Finally, this thesis presents a multiwavelength study of a different type of radio AGN detected in HI with ASKAP-FLASH, demonstrating the gains to be made by multi-wavelength follow up of HI detections, and revealing at least three intervening galaxies along that line of sight, one of which is the likely host of the HI. Overall, this thesis reinforces the key role of peaked spectrum sources in studies of active galaxy evolution, jet-gas interactions, and the changing distribution of gas across the different ages of our Universe. Includes publication

Critical point network-based organizational principle of cortical spatiotemporal dynamics

2026-06-04T05:33:37Z

Critical point network-based organizational principle of cortical spatiotemporal dynamics Xu, Yiben Inspired by the field of turbulence and vector field topology in neural activities, this thesis introduces a novel and generalizable organizational principle of cortical spatiotemporal dynamics based on a global network of interacting critical points. Starting from the analysis of human fMRI signals, this thesis highlights the discovery of travelling cortical spiral waves (termed ‘brain spirals’) during both the resting and task states, emphasizing the mechanistic and functional relevance of a novel spiralbased organizational principle of large-scale brain activities. Next, based on human high-density electroencephalography (HdEEG) recordings, this thesis extends the spiral-based organizational principle from wakefulness to non-rapid-eye-movement (NREM) sleep, revealing that this spiralbased organizational principle is also a defining feature of human N2 sleep, which is associated with sleep-dependent-memory-consolidation and aging-related memory decline. Finally, this thesis further expands the spiral-based organizational principle to include other types of critical points (i.e., sinks, sources and saddles), two new recording modalities (Magnetoencephalography/MEG and Electrocorticography/ECoG), and intracranial recordings of non-human primates (marmoset). In two distinct datasets, a global network of interacting critical points can be consistently observed regardless of species, recording modalities and cognitive states. Acting like an organizational skeleton, this network of critical points collectively enables the task-dependent organizations of largescale brain activities, supporting the universal presence of a critical point network-based organizational principle of large-scale brain activities across species.

Greenwashing: Regulatory Enforcement, Prevention and Detection

2026-06-04T02:33:05Z

Greenwashing: Regulatory Enforcement, Prevention and Detection Peng, Shiyao The prevalence of sustainability disclosures is rapidly increasing, with many such disclosures becoming mandatory across major jurisdictions. At the same time, greenwashing – defined as misleading sustainability claims exaggerating or misrepresenting environmental or other sustainability performance – has proliferated. This has attracted intensified global regulatory scrutiny. Despite academic interest, there is limited knowledge about how regulators identify, assess, and sanction greenwashing claims in practice. The thesis addresses this gap, examining greenwashing through a regulatory lens. In so doing, unique greenwashing datasets are systematically constructed, including a dataset of global greenwashing regulatory enforcement cases between 2015-2024, as well as a greenwashing taxonomy consolidated from eight regulatory guidelines. These datasets provide the thesis’ conceptual and empirical foundation, clarifying how greenwashing is characterised in academic research and regulatory action. Building on this foundation, the thesis comprises three interconnected studies investigating: (1) how regulators define, interpret, and act against greenwashing; (2) whether sustainability assurance potentially reduces greenwashing by addressing regulator-relevant subject matters; and (3) how generative large language models (LLMs) can support large-scale automated detection of greenwashing based on regulatory indicators. Includes publication

Carbonate sequestration in the Pacific, Indian and Atlantic oceans over the Cenozoic

2026-06-04T01:55:48Z

Carbonate sequestration in the Pacific, Indian and Atlantic oceans over the Cenozoic Dalvand, Faranak This thesis reconstructs the regional evolution of the carbonate compensation depth (CCD) across the Pacific and Indian oceans during the Cenozoic. Using lithological, carbonate, and paleo-depth data from DSDP, ODP, and IODP drill sites, regional CCD variability was modelled through time. The results reveal strong spatial differences linked to ocean circulation, tectonic gateway changes, Antarctic glaciation, atmospheric CO₂ fluctuations, and climate transitions. The Pacific Ocean records major Neogene CCD fluctuations associated with gateway reorganisations and the late Miocene biogenic bloom, while the Indian Ocean shows significant variability related to Indo-Pacific circulation and monsoon intensification. This thesis also presents the first basin-specific synthesis of carbonate accumulation rates across the Atlantic, Pacific, and Indian oceans since the Cretaceous, highlighting major inter-basin differences in pelagic carbonate burial. The findings demonstrate that long-term carbonate preservation and oceanic carbon storage were strongly controlled by tectonic evolution, deep-water circulation, and global climate change, providing new insights into the evolution of Earth’s carbon cycle and climate system through geological time. Includes publication

Rebuilding Public Trust in Social Media Platforms? A Case Study of Meta’s Oversight Board

2026-06-03T04:46:22Z

Rebuilding Public Trust in Social Media Platforms? A Case Study of Meta’s Oversight Board Cao, Rumeng Social media platforms have become an integral part of everyday life. Power within the digital environment has become concentrated in a few dominant platforms because of user behaviour and platform design. This concentration allows these platforms to exert a strong influence on civic discourse and the public sphere. However, the growing prevalence of misinformation has transformed how public discussions unfold online. Companies, such as Meta, have faced ongoing criticism for permitting harmful content to spread and for their misuse of user data. These companies’ ineffective responses to these issues have further undermined public trust. Trust plays a crucial role in the digital era. It determines whether users remain active and engaged on social media platforms. To restore public trust, Meta established the Oversight Board as an independent body to review user appeals for removing their content and to enhance transparency and accountability in content moderation. This thesis investigates the effectiveness of Meta’s Oversight Board and examines its decision-making processes. Following the platform regulation triangle theory, this study employs the methodology of document analysis and semi-structured interviews. This study analyses Facebook’s regulatory history from 2010 to 2020 and evaluates Meta’s current strategies for addressing misinformation. This study’s findings contribute to broader debates on platform regulation and offer practical insights for other social media platforms seeking to improve content governance.

Investigating the use of steroids in children with Infantile Epileptic Spasms Syndrome: A multi-omics evaluation of gene and epigenetic regulation

2026-06-03T02:11:10Z

Investigating the use of steroids in children with Infantile Epileptic Spasms Syndrome: A multi-omics evaluation of gene and epigenetic regulation Innes, Emily Amy Objectives: Infantile epileptic spasms syndrome (IESS) is a severe developmental and epileptic encephalopathy of infancy. High-dose oral prednisolone may induce epileptic spasm remission yet long-term outcomes remain poor and it remains unclear whether steroids are targeting disease mechanisms. We investigated the biological pathways underlying IESS and how prednisolone treatment exerts an effect in IESS. Methods: A multi-omics analyses compared blood samples from infants with IESS at baseline (n=11) to controls (n=11) and IESS pre-post prednisolone treatment (n=11). Analyses included bulk RNA sequencing, proteomics, phosphoproteomics and neuroinflammation panel testing. Pathway enrichment analysis using GSEA and ORA identified significantly enriched pathways based on FDR-adjusted p-values. Results: Infants with unknown aetiology had better developmental outcomes than infants with structural aetiologies. Prednisolone induced significant leukophilia, neutrophilia and lymphopenia (all adjusted p-value <0.05). BDNF was significantly elevated at baseline, and prednisolone caused significant increase in nerve growth factor, and significant decrease in the chemokine CCL2. Using RNA seq, prednisolone reversed baseline upregulated ribosomal and mitochondrial pathways, and reversed baseline downregulated immune and membrane transport pathways. Using concordance of quantitative RNA and protein to explore the effects of prednisolone, the most upregulated pathway was 'secretory granule membrane' and the most downregulated pathway was 'ribonucleoprotein complex biogenesis'. Phosphoproteomics revealed the most dysregulated pathway at baseline and after prednisolone was 'chromatin binding'. Significance: Altered gene and epigenetic regulation may be an aetiological mechanism underlying IESS. Prednisolone treatment may control epileptic spasms by altering gene expression through immune-mediated, ribosomal and chromatin pathways.

Compassion Satisfaction and Compassion Fatigue in Australian Rural and Remote Rehabilitation Healthcare Workers

2026-06-03T01:16:44Z

Compassion Satisfaction and Compassion Fatigue in Australian Rural and Remote Rehabilitation Healthcare Workers McGrath, Kelly Lucinda Background. Australian rural and remote rehabilitation healthcare workers operate within insurance-based frameworks with standardised KPIs that do not consider the complexities of rural and remote practice. They face isolation, travel, limited resources, and hazardous conditions that affect professional quality of life, including compassion satisfaction (CS) and compassion fatigue (CF), which comprises burnout and secondary traumatic stress (STS). Aim. To examine, for the first time, levels, experiences, risk and protective factors of CS and CF in this work cohort. Methods. The Professional Quality of life (ProQOL) model guided three studies: a scoping review (n=12 studies), semi‑structured interviews (n=16), and a national mixed‑methods survey that included the ProQoL5 scale (n=29). Each informed the next study. Volunteer participants were rural and remote rehabilitation healthcare workers and registered members of their professional body. Results. No studies specific to rural and remote rehabilitation healthcare workers were found by the scoping review; research focused on medicine and nursing, where CS, CF, and burnout were linked to negative work and environmental factors. Interviews revealed that poor support and safety cultures normalised WHS risks. Survey findings showed lower CS, higher burnout and worse STS than mostly urban Australian healthcare workers Organisational impacts included poor work-life balance and work culture. Conclusions. Rural and remote rehabilitation healthcare workers may experience lower CS and higher CF than urban colleagues. Reported organisational factors align with psychosocial hazards identified in Safe Work Australia legislative updates. The ProQOL5 scale may not fully capture these hazards and therefore needs to be validated in this cohort. Addressing organisational conditions through supervision and workload management is important for workforce sustainability. Includes publication

Memory, Perception & the Art of Seeing Double

2026-06-02T04:34:39Z

Memory, Perception & the Art of Seeing Double Brollo, Deidre This project seeks to examine the role of memory in the viewing of art, in a thesis, ‘Memory, Perception & the Art of Seeing Double’, and an exhibition, ‘the return room’. Drawing on the writings of Henri Bergson and Marcel Proust, the discussion argues for the interrelationship of perception and memory. In contending that perception is inherently bound up with memory, it argues that all artworks have the potential to elicit acts of recollection in the viewer. These acts of recollection should not be understood as an individual reverie detached from the world, but rather as a sort of introspective engagement, an act of viewing that encompasses both the present moment and a past one, in which individual recollection is brought to bear upon the artwork. It is therefore a process of seeing double. Subsequent chapters investigate the use of forms of memory technology in art, looking firstly at the privileged though problematic forms of the photograph and the public memorial. This is followed by an examination of text, maps, and books. The exhibition ‘the return room’ consists of an installation of artists’ books in a specially constructed room reminiscent of both a domestic and a gallery space. It draws on the idea of the palimpsest as a site in which other traces are visible, while also having the potential to be read as a form of memory theatre. The exhibition employs the forms of memory technology discussed in the thesis, with text, maps, and photographs, used within the books. This project, then, claims to be original in the following ways. It claims that Proust and Bergson’s demonstration of the interrelation of memory and perception has important implications for the role of memory in the viewing of all artworks. Secondly, this project draws attention to the possibilities of other forms of memory technology (text and maps) which have not been previously discussed in an art context with regard to their mnemonic potential in relation to the viewer. Thirdly, emphasising that memory is about the juxtaposition of past and present moments, the discussion here offers the models of the palimpsest and the stereoscope as useful conceptual models to describe the ‘art of seeing double.’ Finally, in taking a broad approach to the concept of the artist’s book, the project suggests that the idea of the book may move beyond its boundaries, and that the idea of the palimpsest can be applied in an extended sense to installation space. Includes publication

Mathematical Models for Immune Checkpoint Blockade and Delayed Responses

2026-06-02T03:09:36Z

Mathematical Models for Immune Checkpoint Blockade and Delayed Responses Zheng, Collin Yarmeng Immune checkpoint blockades have transformed oncology, yet a persistent clinical puzzle remains: Why do some patients exhibit delayed responses, with tumours that initially grow or plateau before abruptly regressing? This thesis tackles that question with a multi-scale mathematical study that couples analytically tractable ordinary differential equation (ODE) models with a spatial, stochastic agent-based model (ABM). In ODE form, we show that delayed responses can arise intrinsically, without imposed time lags, via costimulation bottlenecks and slow passages near tipping points associated with special saddle-node bifurcations. We map delayed responses to a statistically-thin part of the model parameter space, suggesting their rarity. Our ODE results enable us to propose an immune profile framework that maps patient prognosis to the natural strength of their immune system---an idea that has become increasingly popular in clinical research since COVID-19. To move beyond mean-field assumptions, we develop an ABM tracking cancer cells, dendritic cells (DCs), CD8+ T cells, and Tregs at single-cell resolution, with molecular attributes and cell-level rules. Mechanistically, our ABM explains why combination therapy outperforms monotherapy: anti-CTLA-4 'reopens the gate' while anti-PD-1 'lifts the brake', yielding a larger and fitter effector CD8+ T cell pool. Our results supports two hypothesised mechanisms of action underlying CTLA-4 blockades---Treg depletion in humans and Treg-driven stripping of B7 ligands---highlighting how depleting suppressors and protecting strategically-important ligands reopen the costimulatory pathway. We characterise delayed responses as an alignment of a multitude of immune events, followed by a fast cascade of killing. This suggests that DC therapies prioritising net DC recruitment and T cell therapies that prioritise tumour-infiltrating lymphocyte (TIL) survivability synergise well with immune checkpoint blockades.

Animating the “Outside”: a Tripartite Model of Analysing 1960s Jazz

2026-06-01T05:58:39Z

Animating the “Outside”: a Tripartite Model of Analysing 1960s Jazz Clarkson, Timothy Nicholas Garrett The revolutionary change in jazz improvisation of the 1960s featured a sudden increase in the degree and duration of departure from pre-composed forms. This practice, commonly known as “outside” playing, has generally been interpreted in musicology and music theory through the lens of “dissonance.” Black radical scholars claim a different, Afrological ontology of dissonance, distinct from the Eurological ontology and philosophy that underpins most mainstream music theory. In this thesis, I argue that a music-theoretical focus on outside playing’s technical dimension has produced a Eurological attunement in discussions of 1960s jazz, neglecting its interactive dimension and cultural practice. Benjamin Givan’s alternative conception of “apart” playing foregrounds interactivity through layers of relationships between musicians working in a group dialogue with composed materials (the “referent”). Black radical discourse more strongly foregrounds the cooperative togetherness that “apart” playing requires, and resists the necessity for resolution and unity. I adopt Fred Moten’s use of “appositional” playing to reflect both these dimensions of improvisational practice and its cultural resonances. I develop new animated music-theoretical tools that innovatively redeploy the Neo-Riemannian Tonnetz to illuminate the dynamic nature of “appositional” playing outlined above. My rationale is anchored in Eric Isaacson’s argument for advantages of animation over still images in engaging temporal relationships. Two case studies investigate strategies within John Coltrane and Ornette Coleman’s ensembles for problematising the referent, developing evidence supporting the sonic phenomena listeners have regularly identified as “transcendence” and “freedom” in their music. These case studies demonstrate the unique advantage of animated tools for investigating the technical and interactive layers of “appositional” playing, and for tying this evidence to the music’s socio-cultural moment.

A multi-point maximum principle to prove global parabolic Harnack inequalities

2026-06-01T03:30:51Z

A multi-point maximum principle to prove global parabolic Harnack inequalities Slegers, Jessica Rachel In this work, we introduce a novel methodology for proving global pointwise Harnack inequalities for parabolic partial differential equations on a Riemannian manifold. The main idea of our approach is to apply a multi-point maximum principle. We demonstrate our techniques by studying the Harnack inequalities satisfied by positive solutions of the linear Schrödinger equation and the doubly nonlinear heat equation. In Chapter 1, we recount the history of parabolic Harnack inequalities, before reviewing the existence of solutions to our aforementioned equations of interest in Chapter 2. In Chapter 3, we present the first proofs of the Harnack inequality using our multi-point maximum principle approach, focusing on classical solutions. In Section 3.1, we analyse the Schrödinger equation, first in Euclidean space and then on a Riemannian manifold with nonnegative Ricci curvature. This section contains applications to Schrödinger equations with a gradient drift term, including the heat equation governed by the Ornstein-Uhlenbeck operator. In addition, we use our Harnack inequality to recover a differential Harnack inequality comparable to the famous result of Li and Yau. In Section 3.2, we demonstrate how our techniques can be extended to prove the Harnack inequality for positive classical solutions of the doubly nonlinear heat equation. However, since solutions of this equation do not in general possess sufficient regularity to be treated as classical solutions, we dedicate Chapter 4 to adapting our proof techniques to viscosity solutions. After reviewing the basic notions associated with viscosity solutions, we develop a modified version of the parabolic theorem on sums by Crandall and Ishii, which is crucial in our methodology. Finally, we present a new proof of the Harnack inequality satisfied by positive viscosity solutions of the doubly nonlinear heat equation.

Efficient and Robust Self-Supervised Learning for Deep Learning-Based Healthcare Applications

2026-06-01T02:43:00Z

Efficient and Robust Self-Supervised Learning for Deep Learning-Based Healthcare Applications Wang, Hao As healthcare increasingly relies on deep learning for medical imaging, a critical challenge arises: the scarcity of labeled data due to expensive and time-consuming manual clinical annotation. This thesis addresses the mismatch between deep learning's heavy data demands and clinical data scarcity by exploring Self-Supervised Learning (SSL). SSL learns meaningful representations from unlabeled data, significantly reducing dependency on extensive annotations by leveraging inherent data structures and relationships. The primary objective of this research is to develop novel SSL methodologies tailored to distinct healthcare analysis tasks, maximizing the efficient use of limited data across multiple scales and modalities. This is demonstrated through three domain-specific innovations: 1. Histopathology: A novel SSL framework leverages the multi-resolution nature of whole-slide images to enable hierarchical representation learning. This effectively captures both global tissue organization and fine-grained cellular details. 2. Dermatology: To mirror clinical workflows, SSL is customized with pretext tasks that align multi-modal representations (clinical and dermoscopic images) and encode inter-label dependencies for complex diagnostic predictions. 3. Remote Physiological Measurement: To extract subtle spatiotemporal signals from facial videos, SSL is extended with physiology-aware temporal and spatial augmentations. This preserves periodic signal integrity while efficiently suppressing noise. Through these investigations, this thesis demonstrates that SSL can be successfully adapted to exploit domain-specific data characteristics—such as multi-resolution hierarchies, multi-modal complementarity, and spatiotemporal dynamics. Ultimately, this research introduces a robust, general SSL framework that significantly reduces annotation requirements while consistently achieving state-of-the-art predictive performance across diverse healthcare applications. Includes publication

Contact and non-contact non-destructive detection of debonding of tiles

2026-06-01T02:26:30Z

Contact and non-contact non-destructive detection of debonding of tiles Zhao, Yu Debonding of tiles in high-rise buildings poses safety threats, making reliable non-destructive inspection (NDI) essential. This study systematically investigates contact and non-contact acoustic NDI methods for detecting tile debonding or loose fixings, elaborating on their mechanisms and procedures. It also provides algorithms applicable to practice, such as identifying loose slot wedges in hydropower generator stators. First, a non-contact acoustic method using a directional sound source and Laser Doppler Vibrometer (LDV) is established. A numerical model simulates acoustic interaction with tiles. The debonding area is identified by plotting the out-of-surface velocity map based on vibration amplitudes at resonance frequency. Numerical simulations agree with experiments, confirming accuracy for various debonding shapes. To enhance efficiency, a deep learning (DL) method is proposed. Continuous wavelet transform converts signals into time-frequency scalograms to build a signature database. Two DL networks are trained: the first classifies debonding types (100% accuracy with a single scalogram), and the second identifies unknown shapes (errors 1–31%). This two-stage method offers a fast, effective solution. The non-contact approach is extended to inspect slot wedges in hydropower generators. Wedges are classified as loose, intermediate, or normal. Using a directional sound source and analyzing frequency peaks (900–2100 Hz), loose and normal conditions achieve 100% accuracy; with DL, all three reach 100%. This method reduces operation time compared to contact techniques. Finally, a contact inspection method using a hammer and microphone mimics human hearing. Debonded areas produce distinct acoustic waveforms. A Fast Fourier Transform-based approach identifies resonance frequencies, and the Digital Damage Fingerprints method maps debonding. This eliminates subjectivity inherent in worker-dependent methods, yielding more reliable results

Computational modelling of spatial contagion dynamics: epidemics, infodemics and socio-economic turbulence

2026-06-01T02:14:37Z

Computational modelling of spatial contagion dynamics: epidemics, infodemics and socio-economic turbulence Jamerlan, Ma Christina Contagions, ranging from epidemics to infodemics and socio-economic turbulence, are often studied in isolation despite exhibiting analogous spatiotemporal transmission dynamics. In this thesis, we develop a unifying multi-city framework for modelling spatial contagions, integrating contagion dynamics, risk disposition, population mobility, and resource distribution. By extending classical multi-city epidemic models, we introduce dynamically adaptive risk-driven mobility flows and examine how population mobility, susceptibility acquisition, risk disposition and effectiveness of distributed resources jointly shape contagion severity and resultant spatial patterns across multiple contagion types. Our results show that small changes in risk disposition or resource effectiveness parameters can lead to substantial shifts in contagion dynamics, revealing phase transitions and tipping points in resultant contagion patterns. We introduce a novel metric, the average cluster intensity, to quantify mean contagion cluster intensity and measure emergent phenomena, such as shield immunity. In some contagion types, altruistic interactions between inoculated and affected individuals reduce overall contagion severity and fragment spatial spread. This shielding effect is most pronounced in socio-economic turbulence, moderate in epidemics, limited in social myth spreading, and not observed in polarisation dynamics. Our case studies using Australian data on COVID-19 incidence, crime records, conflict exposure during protests, and real estate activity confirm that Turing-like patterns are observed empirically in concordance with our model's predictions. Overall, this thesis provides a robust framework for understanding how risk disposition, susceptibility, mobility, and resource distribution collectively drive spatial contagion dynamics. Findings may guide policymakers in designing interventions, allocating resources, and mitigating contagion impacts across diverse societal domains.

Automatic Privacy Compliance Checks for Mobile Apps Using Natural Language Processing

2026-06-01T01:23:15Z

Automatic Privacy Compliance Checks for Mobile Apps Using Natural Language Processing Pinchahewage, Bhanuka Malith Silva The rapid growth of the mobile app ecosystem has intensified concerns about how user data is collected, shared, and communicated through privacy disclosures. Privacy compliance in app marketplaces relies heavily on developer self-reporting and user awareness. As a result, privacy information, whether in detailed policy documents or in summarised forms, often fails to accurately reflect intended data practices. This thesis explores how recent advances in NLP can enable automated and scalable privacy compliance checks in the Google Play Store. It identifies key factors that limit the transparency and usability of privacy policies and proposes enhanced parsing and structuring techniques to improve comprehension and support more effective regulatory oversight. Existing encoder-based models provide accurate predictions but lack interpretability, while decoder-based LLMs provide meaningful explanations, yet they lack verifiability. To address this gap, this thesis first introduces an entailment-driven LLM framework that couples generative reasoning and re-evaluation strategies with embedding-based verification, improving both the interpretability and factual consistency of privacy policy classification. It then presents PrivPRISM, a novel language-modelling framework that leverages both encoder and decoder architectures for large-scale compliance analysis, which cross-examines privacy policies, Play Store disclosures, and installation artefacts to detect inconsistencies. Findings reveal that 53% of analysed apps exhibit discrepancies, highlighting the need for evidence-driven auditing. Finally, this thesis details PrivSTRUCT, a structured modelling approach that leverages developer-defined structural cues to disentangle complex privacy disclosures by linking data items to their stated or implied purposes. The findings reveal a persistent transparency gap in which broadly defined purpose disclosures obscure sensitive first- and third-party data practices in mobile apps.

Automated Mobile Content Compliance Verification Using Multimodal Learning

2026-06-01T01:08:33Z

Automated Mobile Content Compliance Verification Using Multimodal Learning Denipitiyage, Dishanika Dewani The rapid expansion of the mobile app ecosystem has intensified concerns about exposure to inappropriate or misleading content, particularly for children. Although regulatory frameworks such as the GDPR, and app store policies aim to standardise age-appropriate content, mobile marketplaces still rely heavily on developer-declared ratings. Consequently, content rating compliance remains largely underexplored compared to privacy, security, and malware detection. Investigating the detection of content rating non-compliance in mobile apps, this thesis first introduces a multimodal similarity search pipeline to identify app metamorphosis, capturing substantial app evolution over five years. By combining text and visual embeddings with a majority-voting correspondence strategy, the study quantifies app progression and reveals the prevalence of rating inconsistencies in the Google Play. Second, the thesis proposes a vision–language representation learning framework that jointly analyses app descriptions and visual creatives to detect rating violations, leveraging a cross-attention module to align textual and visual semantics, while ListMLE loss models the ordinal structure of content ratings. Next, addresses cross-platform rating inconsistencies by leveraging the Apple App Store as a reference. A content-descriptor-driven data generation pipeline converts app creatives and descriptions into structured question–answer pairs, enabling interpretable descriptor-level prediction using a vision–language model. A two-stage training strategy combining supervised fine-tuning and mistake-driven preference optimisation significantly improves recall over baseline models, enabling cross-platform content compliance auditing in mobile app ecosystems. Building on this ordinal modelling, the thesis concludes with RankOOD, a unified framework that detects out-of-distribution samples by analysing class-wise ranking violations in model outputs, achieving state-of-the-art performance.

Targeted evidence-based care in bronchiectasis in a regional centre: a treatable traits approach to improving clinical and implementation outcomes

2026-05-31T23:48:12Z

Targeted evidence-based care in bronchiectasis in a regional centre: a treatable traits approach to improving clinical and implementation outcomes Krieg, Kirsty Elise Bronchiectasis is a syndrome that develops from a complex interaction of pathophysiological mechanisms, where permanent, abnormal airway dilation is the defining feature. Symptoms, recurrent exacerbations and hospitalisation are key factors determining the severity of bronchiectasis. Exacerbations impact health-related quality of life and disease progression, with frequent exacerbations associated with a higher risk of hospitalisation and mortality. National and international guidelines in bronchiectasis outline the current evidence-based interventions for bronchiectasis. However, it is increasingly recognised that the co-existence of bronchiectasis with other respiratory diseases such as chronic obstructive pulmonary disease (COPD) and asthma, along with the large number of possible comorbid conditions, adds further complexity to management. It is difficult for management guidelines to address the individualised application required, in the presence of complex and unique clinical presentations. Perhaps due to these factors, adherence to guideline recommendations is reported as low in bronchiectasis, resulting in sub-optimal treatment outcomes. New treatment approaches have been proposed in other chronic respiratory diseases, which account for the heterogeneity of clinical presentations. Such approaches are focused on identifying treatable targets or traits of respiratory disease through a structured assessment, and the prioritisation of traits for treatment together with the patient. While treatable traits have been described in bronchiectasis, the approach has not yet been implemented and evaluated in clinical practice. With the known clinical heterogeneity and reported low adherence to guideline-informed care, it is important to test approaches that further individualise care. The treatable traits approach may offer a model that can improve health outcomes in people with bronchiectasis, with interventions that are guided by the patients’ priorities. Includes publication

Advancing Additive Manufacturing of Copper Alloys: Processing, Microstructure, and Property Optimisation

2026-05-29T02:59:25Z

Advancing Additive Manufacturing of Copper Alloys: Processing, Microstructure, and Property Optimisation Chen, Kangwei Copper (Cu) and its alloys are indispensable to modern society due to their exceptional electrical and thermal conductivity, mechanical performance, and corrosion resistance. The transition towards Industry 4.0 and beyond has intensified demand for advanced Cu-based materials. Additive manufacturing (AM) offers the potential to realise these requirements through design flexibility, reduced material waste, and component customisation. However, its application to Cu alloys remains hindered by challenges intrinsic to Cu, such as high reflectivity and rapid heat dissipation. AM imposes cyclic, spatially localised energy inputs that generate steep thermal and stress transients, producing microstructural phenomena not predicted by steady-state metallurgy. Consequently, the fundamental links between powder feedstock, processing conditions, microstructural evolution, post-processing and the resulting mechanical and functional properties are not yet well understood, limiting the widespread adoption of AM Cu alloys. This thesis systematically investigates how AM process parameters, alloying strategies, and powder feedstock characteristics govern the microstructure and performance of three representative Cu alloys—Cu-10Sn, Cu-1Ti, and Cu30Ni. Through a combination of advanced microscopy, mechanical and electrical testing, computational fluid dynamics simulations, thermodynamic simulations, and density functional theory calculations, this thesis establishes quantitative links between processing conditions, microstructural features, and macroscopic properties. Collectively, the findings provide new insights into the solidification pathways, microstructural evolution, and strengthening mechanisms unique to AM Cu-based alloys and deliver practical guidelines for optimising alloy and process design. By bridging fundamental metallurgy with AM-specific processing, the thesis contributes to enabling Cu alloys as next-generation functional and structural materials.

Use of the Sonomat for Evaluating Nocturnal Body Movements in Children.

2026-05-28T13:07:00Z

Use of the Sonomat for Evaluating Nocturnal Body Movements in Children. Lu, Mimi Han Qing Movements during sleep are routinely observed but not consistently quantified in paediatric sleep assessment. This thesis evaluates the Sonomat (MAT) alongside polysomnography (PSG) for measuring sleep-related body movements. In a retrospective cohort of children with concurrent Sonomat and PSG studies, movements were scored using event-duration thresholds (≥1s, ≥3s, ≥5s, ≥7s). Movement index (MI, events/h) and movement duration (MD, % of time) were determined. Analyses examined inter-system agreement, automated scoring, and whether movement burden differed by obstructive sleep apnoea (OSA) status, including within children with McGill oximetry scores of 1. The Sonomat consistently measured higher MI and MD than PSG, detecting more brief movements. This difference lost statistical significance for MI when restricted to movements ≥7s. MD remained statistically significant, though the clinical relevance of a 0.9% difference (~4.5 min of median total sleep period) is unclear. At the 3s threshold, inter-system agreement reached 88%. MD emerged as the preferred burden metric, being less sensitive to event-splitting or merging than MI. Automated MAT scoring showed asymmetry, with MD most closely approximating manual scoring. Movement burden, especially MD, strongly correlated with wakefulness. No discrimination by OSA status was found in overall or sleep-restricted analyses. Small cohort size and retrospective design limited power for sub-analyses, including within the McGill score 1 group. Snoring and stertor are captured by Sonomat but not by the mixed obstructive apnoea-hypopnoea index (MOAHI). In summary, the Sonomat is viable for measuring sleep-related body movements, detecting more brief events than PSG but converging for events ≥7s. MD provides a more robust burden measure than MI. Movement metrics tracked wakefulness but did not differentiate OSA status by MOAHI. Future work should examine larger cohorts and broader sleep-disordered breathing criteria.

Understanding treatment needs to improve care and outcomes for children and adults with rheumatic conditions and their caregivers.

2026-05-28T12:23:55Z

Understanding treatment needs to improve care and outcomes for children and adults with rheumatic conditions and their caregivers. Kelly, Amy Helen The importance of involving the patient in making decisions about their health care and in turn for clinicians to understand the patient’s perspective is recognised. Building on the existing literature, within this thesis a systematic review of outcome measures reported in myositis randomised control trials was conducted and identified that the majority of outcomes reported were surrogate markers and there were few patient reported outcome measures (PROMs). It was also identified that there was very limited data about patient and caregiver experiences in Juvenile Dermatomyositis (JDM) research and to further investigate this, two qualitative research projects were conducted. The first explains the experiences and perspectives of parents who have children diagnosed with JDM and the second study examines parents’ perspectives on the outcome measures important to them. Part 2 of this thesis evolved after the COVID-19 pandemic to investigate the current landscape of telemedicine in Australia and the patient’s perspective of utilising these services. A narrative review was conducted examining how telemedicine is utilised in health care in Australia and the benefits and the disadvantages, in the management of chronic disease and more specifically rheumatic diseases. A qualitative study was then carried out, investigating the experiences of rheumatic disease patients in a metropolitan centre, that provides insights into the patient’s perspective when using telemedicine, to help inform clinicians and administrators as to how to best use this modality to improve health outcomes. Includes publication

Psychological injury: A quantitative assessment of natural justice and the optimum management of psychological factors in compensations systems

2026-05-28T11:46:01Z

Psychological injury: A quantitative assessment of natural justice and the optimum management of psychological factors in compensations systems McMahon, John Edward This thesis documents the creation of psychosocial support program and interdisciplinary clinics for people with insured injuries. Applying machine learning and artificial intelligence, insights are derived from eight single arm studies, including recovery pathways, feigning spectrum behaviour, and the impact of interventions for different insured injuries. A comprehensive narrative review shows the evolution of language from an instrument for cooperation, the means of incorporation, trauma, and the recent development of Generative Artificial Intelligence as language incarnate. Psychological injuries are elucidated. There is a review of literature showing how stakeholder interactions can impact recovery from injury and the need for a trauma informed care approach. The predictive value of verbal and non-verbal expressions of psychological distress on recovery are demonstrated through the application of the Manchester Colour Wheel to a cohort of 1098 injured workers. Machine learning models to compare recovery from work related shoulder injury and motor crash related whiplash, demonstrates the diverse factors in recovery from insured injury. Machine learning models were used to identify the significant psychosocial factors important to the vexing and costly problem of clinical non-attendance. Cut scores for simulation were determined for some common psychometric measures. Large Language Models were used to derive insights from more than 7472 injured workers using a new approach called "persona generation". So called "thinking" large language models generated recovery personas in 711 motor accident injured people. Time series analysis was used to show the locus of natural justice is not with laws per se but at the case manager or business unit level within compensation systems. Detailed recommendations were made for applying trauma informed care and artificial intelligence to maximise natural justice and improve the recovery journeys of people with insured injuries. Includes publication

Using Multi-omics To Study 2-Hydroxyglutarate (Patho)Biology

2026-05-28T10:05:39Z

Using Multi-omics To Study 2-Hydroxyglutarate (Patho)Biology Vigder, Niv The term 2-hydroxyglutarate (2HG) is often used broadly in the biomedical literature, yet it overlooks a key biochemical nuance: chirality. 2HG exists as two enantiomers, L2HG and D2HG, which are structurally identical except for the configuration around the chiral hydroxyl-bearing carbon at the C2 position. This enantioselectivity is biologically consequential as the two forms arise from distinct metabolic pathways engaged under different physiological and pathological stresses. L2HG, but not D2HG, accumulates robustly under conditions of hypoxia, acidosis, and myocardial ischemia. This thesis begins with an introduction (Chapter 1) tracing the evolution of 2HG research from its early chemical characterization to its recognition as a signaling metabolite. Building on this conceptual framework, the first results chapter (Chapter 2) investigates the relationship between L2HG and triglyceride / fatty acid metabolism and demonstrates that conditions promoting L2HG accumulation are accompanied by coordinated remodeling of neutral lipid pools, consistent with altered fatty acid handling and energy storage pathways. The second results chapter (Chapter 3) focuses on phosphatidylethanolamine metabolism, implicating L2HG in regulation of glycerophospholipids. The final results chapter (Chapter 4) moves beyond reductionist analysis by applying a holistic, network medicine–based framework to integrate proteomics data, leading to the identification of major vault protein (MVP), the principal structural component of vault nanoparticles, as a previously unrecognized molecular target of L2HG-associated metabolic stress. Finally, Chapter 5 integrates the findings of this thesis and highlights future directions for the field. Collectively, this thesis defines a connection between L2HG metabolism and lipid and protein remodeling, establishing an integrated framework for understanding how L2HG functions as a metabolic signal in (patho)biology. Includes publication

Mobilising Social Capital and Participation: Deaf Organisations and Access in Disasters

2026-05-28T09:37:33Z

Mobilising Social Capital and Participation: Deaf Organisations and Access in Disasters Craig, Leyla Disasters expose inequities in how information and support are designed, delivered, and accessed, particularly for groups whose linguistic, cultural, and access needs are overlooked. Deaf people, at the intersection of disability and cultural-linguistic minority status, remain underserved in disaster risk reduction (DRR) research and practice despite barriers to lifesaving information. Taking Deaf support organisations as its starting point, this thesis examines how they mobilise alongside health and emergency services, challenges in delivering accessible disaster information, and systemic gaps that leave deaf people at heightened risk. It investigates how mobilisation strategies shape access and participation in DRR. Drawing on qualitative case studies from nine countries (Aotearoa New Zealand, Australia, Haiti, Indonesia, Italy, Japan, Nepal, the Philippines, and the USA), this research explores how organisational strategies shape access, participation, and social capital for resilience. Grounded in Deaf ways of knowing, critical intersectionality, and Bourdieu’s concepts of capital, field, and habitus, I use reflexivity to examine my positionality as a deaf researcher and how it shaped the study. Findings show Deaf support organisations play a critical undervalued role in enabling accessible communication during disasters. Despite limited resources, they bridge communication gaps and support resilience, though their role remains under-recognised in emergency management. Exclusion is shaped by systemic barriers and internal hierarchies within Deaf communities, disproportionately affecting deaf people who are further marginalised. Building on this, the thesis develops a relational framework integrating social capital and participatory inclusion to explain how power, recognition, and access shape participation in DRR. It argues that strengthening Deaf support organisations is essential for inclusive disaster preparedness and response. Includes publication

Deep-time reconstructions of Earth's surface environments and elevations

2026-05-28T04:38:14Z

Deep-time reconstructions of Earth's surface environments and elevations Singh, Satyam Pratap Reconstructing Earth's ancient surface topography in deep geological time demands the synthesis of plate tectonic reconstructions, geodynamic simulations, paleoclimate modelling, and advanced computational methodologies. This thesis pioneers an integrated computational framework bridging geological observations with numerical models for paleotopographic reconstruction. A novel deformable plate tectonic reconstruction is developed incorporating time-evolving deforming meshes within rift zones, applied to the Gulf of Mexico. Systematic optimisation across 32,400 mesh configurations reduced crustal-thickness RMSE from 14.8 km to 5.6 km against the GEMMA model. The resulting subsidence histories illuminate key depositional enigmas, including ~1.5 km of pre-drift subsidence during the Sinemurian (193–183 Ma), southward migration of red-bed deposition beneath Jurassic salt, and the westward deflection of Cenomanian–Turonian clastic systems. Transitioning to active margins, the Python Deep Time Data Mining (pyDTDM) library is introduced within an Explainable AI (XAI) framework, integrating plate reconstructions, mantle convection simulations, and paleoclimate outputs. The XAI model identifies subduction flux as the dominant orogenic driver, with trench-advance episodes intensifying crustal thickening, while mantle temperature anomalies and long-term precipitation exert secondary but significant influences. Leveraging these insights, a deep neural network reconstructs active-margin paleotopography at 1 Myr resolution throughout the Mesozoic and Cenozoic. Validated against geochemical paleoelevation proxies and regional studies, the model reveals the East Asian Cordillera exceeding 3 km during the mid-Cretaceous and reproduces established Andean uplift phases. All workflows are disseminated as open-source tools under GPL/LGPL licences, with broader implications extending to mineral exploration, paleoclimate modelling, and biodiversity evolution studies.