Monash University Research Portal
Not a member yet
41460 research outputs found
Sort by
Towards reliable generative AI-driven scaffolding:Reducing hallucinations and enhancing quality in self-regulated learning support
No full textGenerative Artificial Intelligence (GenAI) holds a potential to advance existing educational technologies with capabilities to automatically generate personalised scaffolds that support students’ self-regulated learning (SRL). While advancements in large language models (LLMs) promise improvements in the adaptability and quality of educational technologies for SRL, there remain concerns about the hallucinations in content generated by LLMs, which can compromise both the learning experience and ethical standards. To address these challenges, we proposed GenAI-enabled approaches for evaluating personalised SRL scaffolds before they are presented to students, aiming for reducing hallucinations and improving overall quality of LLM-generated personalised scaffolds. Specifically, two approaches are investigated. The first approach involved developing a multi-agent system approach for reliability evaluation to assess the extent to which LLM-generated scaffolds accurately target relevant SRL processes. The second approach utilised the “LLM-as-a-Judge” technique for quality evaluation that evaluates LLM-generated scaffolds for their helpfulness in supporting students. We constructed evaluation datasets, and compared our results with single-agent LLM systems and machine learning approach baselines. Our findings indicate that the reliability evaluation approach is highly effective and outperforms the baselines, showing almost perfect alignment with human experts’ evaluations. Moreover, both proposed evaluation approaches can be harnessed to effectively reduce hallucinations. Additionally, we identified and discussed bias limitations of the “LLM-as-a-Judge” technique in evaluating LLM-generated scaffolds. We suggest incorporating these approaches into GenAI-powered personalised SRL scaffolding systems to mitigate hallucination issues and improve the overall scaffolding quality.</p
The forgotten meaning of אוֹת
No full textThe Hebrew noun אוֹת is generally understood to refer to a sign, frequently one that is divine or miraculous. This understanding has long informed interpretations of biblical passages and Hebrew inscriptions. The common definition, we argue, is often inapt and fails to account for many instances of the term. From ancient biblical translations to modern scholarship, myriad ad hoc explanations have been suggested for difficult passages featuring this word, none of which are persuasive. We propose that אוֹת has an overlooked constellation of meanings related to proclamations and commitments, which better explains its usage in several biblical passages. Our proposal is further supported by epigraphic and comparative Semitic evidence
Scalable and effective negative sample generation for hyperedge prediction
No full textHypergraphs have demonstrated their superiority in modeling complex systems compared to traditional graphs by directly capturing the interactions among multiple entities. Hyperedge prediction, which aims to predict unobserved potential hyperedges, is a fundamental task in hypergraph analysis. A critical component in hyperedge prediction is the sampling of informative negative hyperedges from significantly larger candidate negative sets, compared to traditional graphs, to enhance model training efficacy. Most existing methods utilize predefined heuristics to sample negative hyperedges, resulting in limited generalizability due to their reliance on these predefined rules. The new state-of-the-art in this field is generation-based methods, which treat negative sampling as a generative task. Nevertheless, current generation-based approaches are not scalable to large hypergraphs. Additionally, diffusion models have demonstrated superior performance in numerous generative tasks, yet their potential application in the generation of negative hyperedges remains unexplored. However, the adaptation of diffusion models to this specific task presents challenges due to: (1) diffusion models are inherently designed to generate high-quality positive samples, which are well-defined, as opposed to negative samples; (2) diffusion models are traditionally employed in continuous space, whereas negative sampling for hyperedge prediction operates in discrete space.To address these complexities, we introduce SEHP (Scalable and Effective Negative Sample Generation for Hyperedge Prediction), which employs a conditional diffusion model to iteratively generate and refine negative hyperedges, thereby advancing them towards the decision boundary to improve model performance. SEHP further enhances scalability by effectively sampling sub-hypergraphs, integrating global structural information into the diffusion model for batch training. Extensive experiments conducted on real-world datasets demonstrate that SEHP surpasses existing state-of-the-art methods in both prediction accuracy and scalability. The code of our paper is available at https://github.com/SLQu/SEHP</p
MTS-Net:Dual-enhanced positional multi-head self-attention for 3D CT diagnosis of May-Thurner Syndrome
No full textMay-Thurner Syndrome (MTS) is a vascular condition that affects over 20% of the population and significantly increases the risk of iliofemoral deep venous thrombosis. Accurate and early diagnosis of MTS using computed tomography (CT) remains a clinical challenge due to the subtle anatomical compression and variability across patients. In this paper, we propose MTS-Net, an end-to-end 3D deep learning framework designed to capture spatial–temporal patterns from CT volumes for reliable MTS diagnosis. MTS-Net builds upon 3D ResNet-18 by embedding a novel dual-enhanced positional multi-head self-attention (DEP-MHSA) module into the Transformer encoder of the network's final stages. The proposed DEP-MHSA employs multi-scale convolution and integrates positional embeddings into both attention weights and residual paths, enhancing spatial context preservation, which is crucial for identifying venous compression. To validate our approach, we curate the first publicly available dataset for MTS, MTS-CT, containing over 747 gender-balanced subjects with standard and enhanced CT scans. Experimental results demonstrate that MTS-Net achieves average 0.79 accuracy, 0.84 AUC, and 0.78 F1-score, outperforming baseline models including 3D ResNet, DenseNet-BC, and BabyNet. Our work not only introduces a new diagnostic architecture for MTS but also provides a high-quality benchmark dataset to facilitate future research in automated vascular syndrome detection. We make our code and dataset publicly available at: https://github.com/Nutingnon/MTS_dep_mhsa.</p
“I am rooted, but I flow”:a photovoice investigation into the (re)construction of professional identity among immigrant early childhood educators in Australia
No full textImmigrant ECEC educators in Australia navigate unique personal, cultural, and institutional conditions as they (re)construct their professional identities in a new sociocultural context. This study employs Photovoice and two rounds of in-depth interviews with twelve immigrant ECEC educators to explore the factors that facilitate or impede this identity construction process. Using a “garden elements” metaphor developed through constructivist grounded theory analysis, the findings highlight three key facilitators of identity construction, namely reflective practices that anchor growth (“roots”), community and institutional supports that provide emotional and professional nourishment (“soil”), and everyday acts of creative agency that energise practice (“sunlight”). Conversely, three impediments that disrupt continuity and contribute to moments of fragmentation include emotional and resource scarcity (“drought”), tensions arising from negotiating multiple cultural and professional expectations (“weeds”), and periods of reflective immobility (“pests”). Methodologically, this study extends Photovoice by adapting it to individual interviews and by giving equal analytic weight to visual and narrative data. Implications for supporting immigrant ECEC educators include strengthening culturally responsive induction structures, embedding reflective learning opportunities, and recognising theseeducators’ creative and cultural expertise as central to high-quality ECEC practice.</p
Proposing a multi-level perspective of the influencing factors of repair behaviours of household electrical appliances
No full textWithin the framework of the circular economy, extending the lifespan of electrical products through repairing them is a promising approach, particularly against the backdrop of growing e-waste levels. Consumers play a key role in facilitating repair and multiple studies have suggested their behaviour is influenced by combinations of different internal (e.g. attitudes, knowledge and motivation) and external factors (e.g. product design, market conditions and service availability). Despite this, in empirical research, there is still preference for psychological theories that foreground the individual. Thus, the present study moves beyond these theories and is underpinned by a multi-level perspective – an approach that can systematically organise factors to internal and external contexts. Through qualitative interviews with consumers we identify and then organise the factors that influence repair behaviours of household electrical appliances to a multi-level perspective including micro (individual), meso (household) and macro (repair ecosystems and beyond) levels. The main finding from the interviews and framework development was the identification of 18 influencing factors across four nested levels. The primary contribution is the resulting framework that makes explicit the different internal and external contexts that influence repair behaviours. In particular, the framework highlights the important role of the household-level and makes visible how factors and levels interact to influence repair behaviours. Hence, the framework supports program managers and policymakers to design multi-level interventions. We also offer several opportunities for future research to explore the role of household routines and structure and continue to build an understanding of the interactions between levels and factors.</p
How natural disasters spread conflict
No full textThis paper studies how natural disasters spread conflicts within a network. We first construct a new panel data set that combines geo-referenced information about conflict events and natural disasters, for 5,944 districts in 53 African countries, over the period 1989–2020. Considering natural disasters as exogenous shocks that affect the combatants’ activity in a locality, we find that natural disasters decrease conflict incidence in the affected locality, increase conflict incidence in neighbouring localities, and lead to an overall net increase in conflict incidence. The spatial dispersion of conflict varies by the level of local rent-seeking opportunities and the level of international, post-disaster aid. We then provide a simple theoretical framework that may explain this conflict dispersion pattern. Findings provide important implications for implementing local and aggregate level conflict mitigation policies.</p
Healthcare usage and cost-effectiveness of approach bias modification at 12-months for patients undergoing inpatient withdrawal for alcohol use disorder
No full textIntroduction: There is evidence that approach bias modification (ApBM), a type of computerised cognitive training delivered during inpatient alcohol withdrawal treatment, significantly reduces relapse rate. Our analysis examines, for the first time, whether ApBM is cost-effective compared to sham-trained controls. Methods: Patients at four inpatient withdrawal units were randomized to four daily sessions of ApBM, or sham (control) training. Self-reported data on alcohol use, treatment, and healthcare use was collected over 12-months. We conducted a trial-based cost-effectiveness study of ApBM (versus no ApBM) from a health system perspective. Costs were from relevant Australian 2022 sources. We estimated incremental differences between groups in healthcare costs and abstinence rates using mixed generalised linear models. Results: At 12 months after discharge from the index withdrawal treatment episode, two thirds of participants had accessed acute health care services (i.e., inpatient withdrawal, ambulance, emergency department and hospital inpatient). Results generally indicated non-significant increases in cumulative costs (7743, 201,610, with confidence limits ranging from ApBM being less costly and more effective to more costly and less effective than no ApBM. Conclusion: Although there was evidence of improved abstinence rates in the first 3-months post-discharge, delivering ApBM during acute alcohol withdrawal treatment will not likely generate net benefits over a 1-year period at any willingness-to-pay threshold, due to the continued heavy use of healthcare services in this population. Future research should test whether additional ApBM delivered post-discharge (e.g., via smartphone apps) could extend its relapse prevention effects and ultimately result in cost savings in the long-term.</p
Who stands up to persuade? Voluntary influencers in public support for Pigouvian taxation
No full textWe examine how voters choose to influence others' attitudes toward policy, focusing on the context of Pigouvian taxation. Data from a controlled laboratory experiment show that individuals are generally reluctant to stand up and persuade others. Among those who do, both tax supporters and objectors are equally likely to volunteer—and equally persuasive. As a result, overall negative attitudes toward Pigouvian taxes persist. Moreover, it is the strength of individuals' initial views, rather than an informational advantage, that increases the likelihood of self-nomination as first voters, regardless of the direction of those views. These findings help explain the enduring lack of public support for welfare-enhancing tax policies and suggest avenues for addressing it.</p