Search

The digital transformation in education has created many opportunities but also made it challenging to navigate the growing landscape of digital learning materials. The volume and diversity of learning resources create challenges for both educators and learners to identify and utilize the most relevant resources based on specific learning contexts. In light of this, there is a critical demand for systems capable of effectively connecting different learning materials to support teaching and learning activities and, for that purpose, natural language processing can be used to provide some of the essential building blocks for educational content recommendation systems. Hence, this thesis explores the use of natural language processing techniques for automatically linking and recommending relevant learning resources in the form of textbook content, exercises and curriculum goals. A key question is how to represent diverse learning materials effectively and, to that end, various language models are explored; the obtained representations are then used for measuring semantic textual similarity between learning materials. Learning materials can also be represented based on educational concepts, which is investigated in an ontology-based linking approach. To further enhance the representations and improve linking performance, different language models can be combined and augmented using external knowledge in the form of knowledge graphs and knowledge bases. Beyond approaches based on semantic textual similarity, prompting large language models is explored and a method based on retrieval-augmented generation (RAG) to improve linking performance is proposed. The thesis presents a systematic empirical evaluation of natural language processing techniques for representing and linking digital learning content, spanning different types of learning materials, use cases, and subjects. The results demonstrate the feasibility of unsupervised approaches based on semantic textual simila

Self-regulated learning (SRL) is a cognitive ability with demonstrable significance in facilitating students’ ability to effectively strategize, monitor, and assess their own learning actions. Studies have indicated that a lack of selfregulated learning skills negatively impacts students’ academic performance. Effective data-driven feedback and action recommendations are considered crucial for SRL and significantly influence student learning and performance. However, the task of delivering personalised feedback to every student poses a significant challenge for teachers. Moreover, the task of identifying appropriate learning activities and resources for individualised recommendations poses a significant challenge for teachers, given the large number of students enrolled in most courses. To address these challenges, several studies have examined how learning analytics-based dashboards can support students’ self-regulation. These dashboards offered several visualisations (as feedback) on student success and failure. However, while such feedback may be beneficial, it does not offer insightful information or actionable recommendations to help students improve academically. Explainable artificial intelligence (xAI) approaches have been proposed to explain such feedback and generate insights from predictive models, with a focus on the relevant actions a student needs to take to improve in ongoing courses. Such intelligent activities could be offered to students as data-driven behavioural change recommendations. This thesis offers an xAI-based approach that predicts course performance and computes informative feedback and actionable recommendations to promote student self-regulation. Unlike previous research, this thesis integrates a predictive approach with an xAI approach to analyse and manipulate students’ learning trajectories. The aim is to offer detailed, data-driven actionable feedback to students by providing in-depth insights and explanations for the predictions p

Purpose - Online health question-and-answer (Q&A) forums have developed a new business model whereby listeners (peer patients) can pay to read health information derived from consultations between askers (focal patients) and answerers (physicians). However, research exploring the mechanism behind peer patients' purchase decisions and the specific nature of the information driving these decisions has remained limited. This study aims to develop a theoretical model for understanding how peer patients make such decisions based on limited information, i.e. the first question displayed in each focal patient-physician interaction record, considering argument quality (interrogative form and information details) and source credibility (patient experience of focal patients), including the contingent role of urgency. Design/methodology/approach - The model was tested by text mining 1,960 consultation records from a popular Chinese online health Q&A forum on the Yilu App. These records involved interactions between focal patients and physicians and were purchased by 447,718 peer patients seeking health-related information until this research. Findings - Patient experience embedded in focal patients' questions plays a significant role in inducing peer patients to purchase previous consultation records featuring exchanges between focal patients and physicians; in particular, increasingly detailed information is associated with a reduced probability of making a purchase. When focal patients demonstrate a high level of urgency, the effect of information details is weakened, while the interrogative form is strengthened. Originality/value - The originality of this study lies in its exploration of the monetization mechanism forming the trilateral relationship between askers (focal patients), answerers (physicians) and listeners (peer patients) in the business model “paying to view others' answers” in the online health Q&A forum and the moderating role of urgency in explai

Although the field of Artificial Intelligence in Education (AIEd) has a substantial history as a research domain, never before has the rapid evolution of AI applications in education sparked such prominent public discourse. Given the already rapidly growing AIEd literature base in higher education, now is the time to ensure that the field has a solid research and conceptual grounding. This review of reviews is the first comprehensive meta review to explore the scope and nature of AIEd in higher education (AIHEd) research, by synthesising secondary research (e.g., systematic reviews), indexed in the Web of Science, Scopus, ERIC, EBSCOHost, IEEE Xplore, ScienceDirect and ACM Digital Library, or captured through snowballing in OpenAlex, ResearchGate and Google Scholar. Reviews were included if they synthesised applications of AI solely in formal higher or continuing education, were published in English between 2018 and July 2023, were journal articles or full conference papers, and if they had a method section 66 publications were included for data extraction and synthesis in EPPI Reviewer, which were predominantly systematic reviews (66.7%), published by authors from North America (27.3%), conducted in teams (89.4%) in mostly domestic-only collaborations (71.2%). Findings show that these reviews mostly focused on AIHEd generally (47.0%) or Profiling and Prediction (28.8%) as thematic foci, however key findings indicated a predominance of the use of Adaptive Systems and Personalisation in higher education. Research gaps identified suggest a need for greater ethical, methodological, and contextual considerations within future research, alongside interdisciplinary approaches to AIHEd application. Suggestions are provided to guide future primary and secondary research.

Learning analytics (LA) is an important area of research in technology-enhanced learning that has emerged during the last decade. In earlier years, several systematic reviews have been conducted that focused on the theories behind LA or on empirical studies that utilised LA-based methods to improve learning and teaching processes in higher education. However, to date, there has been no systematic review of papers that have adopted a software perspective to report on the many forms of learning analytics software (LAS) that have been developed, despite these being used more frequently than before in higher education to support learning and teaching processes. To fill this gap, this paper presents a systematic review of LAS with the aim of critically scrutinising the ways in which the use of interactive software in real-world settings may both support students in improving their academic performance and assist teachers in various pedagogical practices. A thematic analysis of 75 articles was conducted, resulting in the identification of three categories of LAS: at-risk student identification software; self-regulation software; and collaborative learning software. For each of these categories, we analysed (i) the embedded functionality; (ii) the stakeholder (teacher and student) for which the functionality is intended; (iii) the analytical and visualisation approaches implemented; and (iv) the limitations of the software that require future attention. Based on the findings of our review, we propose future directions for the development of LAS.

This special issue is a collection of emerging trends and challenges in applying learning-driven approaches to data fabric architectures within the cloud-to-thing continuum. As data generation and processing increasingly occur at the edge, there is a growing need for intelligent, adaptive data management solutions that seamlessly operate across distributed environments. In this special issue, we received research contributions from various groups around the world. We chose the eight most appropriate and novel contributions to include in this special issue. These eight contributions were further categorized into three themes: Data Handling approaches, resource optimization and management, and security and attacks. Additionally, this editorial suggests future research directions that will potentially lead to groundbreaking insights, which could pave the way for a new era of learning techniques in Data Fabric and the Cloud-to-Thing Continuum.

Machine learning methods can be used to reduce the manual workload in exam grading, making it possible for teachers to spend more time on other tasks. However, when it comes to grading exams, fully eliminating manual work is not yet possible even with very accurate automated grading, as any grading mistakes could have significant consequences for the students. Here, the evaluation of an automated grading approach is therefore extended from measuring workload in relation to the accuracy of automated grading, to also measuring the overall workload required to correctly grade a full exam, with and without the support of machine learning. The evaluation was performed during an introductory computer science course with over 400 students. The exam consisted of 64 questions with relatively short answers and a two-step approach for automated grading was applied. First, a subset of answers to the exam questions was manually graded and next used as training data for machine learning models classifying the remaining answers. A number of different strategies for how to select which answers to include in the training data were evaluated. The time spent on different grading actions was measured along with the reduction of effort using clustering of answers and automated scoring. Compared to fully manual grading, the overall reduction of workload was substantial-between 64% and 74%-even with a complete manual review of all classifier output to ensure a fair grading.

Automated fraud detection can assist organisations to safeguard user accounts, a task that is very challenging due to the great sparsity of known fraud transactions. Many approaches in the literature focus on credit card fraud and ignore the growing field of online banking. However, there is a lack of publicly available data for both. The lack of publicly available data hinders the progress of the field and limits the investigation of potential solutions. With this work, we: (a) introduce FraudNLP, the first anonymised, publicly available dataset for online fraud detection, (b) benchmark machine and deep learning methods with multiple evaluation measures, (c) argue that online actions do follow rules similar to natural language and hence can be approached successfully by natural language processing methods.

This paper explores the application of various machine learning techniques for predicting customer churn in the telecommunications sector. We utilized a publicly accessible dataset and implemented several models, including Artificial Neural Networks, Decision Trees, Support Vector Machines, Random Forests, Logistic Regression, and gradient boosting techniques (XGBoost, LightGBM, and CatBoost). To mitigate the challenges posed by imbalanced datasets, we adopted different data sampling strategies, namely SMOTE, SMOTE combined with Tomek Links, and SMOTE combined with Edited Nearest Neighbors. Moreover, hyperparameter tuning was employed to enhance model performance. Our evaluation employed standard metrics, such as Precision, Recall, F1-score, and the Receiver Operating Characteristic Area Under Curve (ROC AUC). In terms of the F1-score metric, CatBoost demonstrates superior performance compared to other machine learning models, achieving an outstanding 93% following the application of Optuna hyperparameter optimization. In the context of the ROC AUC metric, both XGBoost and CatBoost exhibit exceptional performance, recording remarkable scores of 91%. This achievement for XGBoost is attained after implementing a combination of SMOTE with Tomek Links, while CatBoost reaches this level of performance after the application of Optuna hyperparameter optimization.

Agent-based models are particularly suitable to reflect the dynamics of humans, nature, and their interactions, making them a crucial approach for understanding social-ecological systems. The formalisations of human decision-making are central to resulting model behaviours. Despite awareness of the complexity of human behaviour in social-ecological systems research, scholars tend to represent human decision-makers as simplified, perfectly informed rational optimisers, without explicitly considering the fit with decision context. Key reasons are a lacking uptake of social theories and insights. To advance, we need a practice of reflecting, sharing, and inquiring on the justification of the decision model fit with its context. This paper stimulates this practice by 1) supporting the justification of decision model (DM) fit by describing the DM landscape and providing guiding questions; and 2) by supporting researchers in considering alternative DMs through a survey-based impression of modeller practices, and through highlighting DM frontiers as inspiration for future research.

This research is about addressing the need to better understand interaction with conversational user interfaces (CUIs) and how human-technology `conversations' can be improved by drawing on the lessons learned from human-human interaction. It focuses on incorporating abstractions of complex human behaviour, specifically gaze, to enhance interactions with speech agents in conversations. Across four empirical studies, a mix of methods is used to look closely at the interaction between the user and the system. I offer empirical and conceptual contributions for interaction designers and researchers. First, I present a novel speech interface, Tama, which is a gaze-aware speech agent designed to explore the use of gaze in conversational interactions with smart speakers. Second, I present the empirical contributions, that is, the studies that document the interactions with and around speech interfaces, including ongoing, non-system-directed speech. A moment-by-moment analysis of these interactions highlights the opportunities that the gaze offers as a modality to enhance the interaction with the speech agent, as well as the problems and limitations when such a modality is used. The third contribution is a conceptual contribution made by providing perspective on minimal anthropomorphic design. This produces interactions that are not human-like in terms of behaviour but do take advantage of the skills used in human interaction as a key to advancing interactions with speech agents. Based on my research work and contributions, I reflect upon advancing interactions with speech interfaces, focusing on what different technologies can offer and the possibility of taking the next step in designing CUIs. I then discuss the need to bridge the work of different fields (i.e. conversation analysis (CA), human-computer interaction (HCI), and human-robot interaction (HRI)) to combine models and approaches from all these fields in order to guide designers building speech systems. I see thr

Addressing complexity in transportation in cases such as disruptive trends or disaggregated management strategies has become increasingly important. This in turn is resulting in the rising adoption of Agent-Based and Activity-Based modeling. Still, a broad adoption is hindered by the high complexity and computational needs. For example, hundreds of parameters are involved in the calibration of Activity-Based models focused on behavioral theory, to properly frame the required detailed socio-economical characteristics. To address this challenge, this paper presents a novel Bayesian Optimization approach that incorporates a surrogate model defined as an improved Random Forest to automate the calibration process of the behavioral parameters. The presented solution calibrates the largest set of parameters yet, according to the literature, by combining state-of-the-art methods. To the best of the authors' knowledge, this is the first work in which such a high dimensionality is tackled in sequential model-based algorithm configuration theory. The proposed method is tested in the city of Tallinn, Estonia, for which the calibration of 477 behavioral parameters is carried out. The calibration process results in a satisfactory performance for all the major indicators, the OD matrix average mismatch is equal to 15.92 vehicles per day while the error for the overall number of trips is equal to 4%.

Purpose - This study aims to explore the information needs of Chinese parents of children with Autism Spectrum Disorder (ASD) and how these needs evolve as their children develop.Design/methodology/approach - This study collated 17,122 questions regarding raising children with ASD via the Yi Lin website until November 2021.Findings - The information needs of parents of children with ASD were classified into two categories: 1) Cognition-motivation: related to children with ASD; and 2) Affection-motivation: related to their parents. Child development causes the adaptation of information needs of these parents. Within the first three years, nine different topics of these parents' information needs were identified. Major information needs at this stage are as follows: intervention content, intervention methods and pre-diagnosis questions. During the ages of three to six years, there were 13 topics of information needs for parents, focusing on three areas: intervention content, intervention methods and diagnosis and examination. There are eight topics of information needs post six years. Parents are more concerned with the three topics of intervention content, life planning and intervention methods.Originality/value - This novel study indicates the complex and changing information needs of parents of children with ASD in China. It may enhance the understanding of the information needs of these parents at theoretical and practical levels, provide support for them to understand their own information needs and provide a reference for relevant government and social organisations to provide targeted information services for them.

Emerging automated-decision making (ADM) technologies invite scholars to engage with future points in time and contexts that have not yet arisen. This particular state of not knowing yet implies the methodological challenge of examining images of the future and how such images will materialize in practice. In this respect, we ask the following: what are appropriate research methods for studying emerging ADM technologies in education? How do researchers explore sociotechnical practices that are in the making? Guided by these questions, we investigate the increasing adoption of ADM in teachers’ assessment practices. This constitutes a case in point for reflecting on the research methods applied to address the future of assessment in education. In this context, we distinguish between representational methods oriented to recounting past experiences and future(s) methods oriented to making futures. Studying the literature on speculative methods in digital education, we illustrate four categories of future(s)-oriented methods and reflect on their characteristics through a backcasting workshop conducted with teachers. We conclude by discussing the need to reconsider the methodological choices made for studying emerging technologies in critical assessment practices and generate new knowledge on methods able to contribute to alternative imaginaries of automation in education.

A population-level analysis is proposed to address data sparsity when building predictive models for engineering infrastructure. Utilizing an interpretable hierarchical Bayesian approach and operational fleet data, domain expertise is naturally encoded (and appropriately shared) between different subgroups, representing (1) use-type, (2) component, or (3) operating condition. Specifically, domain expertise is exploited to constrain the model via assumptions (and prior distributions) allowing the methodology to automatically share information between similar assets, improving the survival analysis of a truck fleet (15% and 13% increases in predictive log-likelihood of hazard) and power prediction in a wind farm (up to 82% reduction in the standard deviation of maximum output prediction). In each asset management example, a set of correlated functions is learnt over the fleet, in a combined inference, to learn a population model. Parameter estimation is improved when subfleets are allowed to share correlated information at different levels in the hierarchy; the (averaged) reduction in standard deviation for interpretable parameters in the survival analysis is 70%, alongside 32% in wind farm power models. In turn, groups with incomplete data automatically borrow statistical strength from those that are data-rich. The statistical correlations enable knowledge transfer via Bayesian transfer learning, and the correlations can be inspected to inform which assets share information for which effect (i.e., parameter). Successes in both case studies demonstrate the wide applicability in practical infrastructure monitoring, since the approach is naturally adapted between interpretable fleet models of different in situ examples.

Weibull time-to-event recurrent neural networks (WTTE-RNN) is a simple and versatile prognosis algorithm that works by optimising a Weibull survival function using a recurrent neural network. It offers the combined benefits of the sequential nature of the recurrent neural network, and the ability of the Weibull loss function to incorporate censored data. The goal of this paper is to present the first industrial use case of WTTE-RNN for prognosis. Prognosis of turbocharger conditions in a fleet of heavy-duty trucks is presented here, where the condition data used in the case study were recorded as a time series of sparsely sampled histograms. The experiments include comparison of the prediction models trained using data from the entire fleet of trucks vs data from clustered sub-fleets, where it is concluded that clustering is only beneficial as long as the training dataset is large enough for the model to not overfit. Moreover, the censored data from assets that did not fail are also shown to be incorporated while optimising the Weibull loss function and improve prediction performance. Overall, this paper concludes that WTTE-RNN-based failure predictions enable predictive maintenance policies, which are enhanced by identifying the sub-fleets of similar trucks.

Purpose - Students' satisfaction with e-learning outcomes is a vital component in determining the adoption of e-learning systems. Understanding the antecedent factors impacting students' satisfaction with e-learning outcomes helps decision-makers at the higher education to take the necessary actions to enhance the quality of students' performance, especially during the exceptional times, such as the COVID-19 pandemic. Design/methodology/approach - This paper aims to determine and analyse the antecedent factors influencing students' satisfaction with e-learning outcomes during the pandemic. Structural equation modelling (SEM) is used to analysis a proposed research model. Findings - The SEM results show that digital communities in e-learning, information technology (quality and accessibility) and the online course design quality directly influence students' satisfaction with e-learning outcomes. More interestingly, results show that the COVID-19-related factors (1) awareness of the COVID-19, (2) perceived challenges (negatively) and (3) the educational institutions' preparedness also influence e-learning outcomes. Research limitations/implications - The results suggest that the decision-makers at the educational institutions should consider adopting a blended learning system approach to deliver e-learning during the emergency, such as the COVID-19 outbreak situation. Originality/value - This is one of the first studies in higher education context, which seek to identify the antecedent factors that influence students' satisfaction with e-learning outcomes during the COVID-19 pandemic.

Purpose: This systematic literature review aims to elaborate the research progress of information practices to help other researchers attain a more holistic and comprehensive understanding of the field. Design/methodology/approach: Following a systematic review protocol, 123 research articles from nine academic databases were included in the analysis. Findings: Four separate results can be outlined. First, practice theory, social constructionist theory and activity theory are often used as the theoretical basis for the study of information practices. Second, people will engage in specific information practice activities in different external and internal contexts. The external contexts include social and needs contexts. The internal contexts include information source horizons, user's affection and user's cognition. Third, the existing information practice models can be divided into static and dynamic types. These models mainly reflected activities and influencing factors of information practices. Fourth, qualitative methods were the most used in information practice research. Research limitations/implications: The field of information practices is a vast, expanding research field. This research will focus only on a specific section, namely concepts, activities, contexts and models. Researchers could contribute to exploring the concepts, components and mechanisms of information practices by combining theories from various disciplines, such as sociology and behavioral science. Originality/value: This is the first study to reveal the general picture of information practices. It also elaborates the characteristics of people's information practices and shows the potential development direction for future research.

Purpose/Problem: Providing a digital platform for citizens to collaborate with governments to address environmental issues is vital for achieving sustainability goals. However, the e-government literature lacks studies that provide pragmatic solutions to sustainability problems through the design of artifacts, especially in developing countries. To fill the gap, the dissertation aims to design and evaluate a digital government collaborative platform (DGCP) and produce design knowledge to facilitate the collaboration between the government and citizens to address environmental sustainability issues in Sri Lanka. Research methods: The design science research methodology (DSRM) was chosen for this research. Value-sensitive design (VSD) was also adopted throughout the design process to identify and embed human values in the DGCP. The prototype and the final workable DGCP were evaluated by stakeholders following the framework for evaluation in design science research (FEDS). Multiple methods were used in data collection and analysis. Results: The result of the research was a designed and evaluated e-government artifact, the DGCP. By adopting the VSD approach, we first derived fifteen human values that were critical to DGCP design. The values were translated into thirty-one design requirements that were integrated into the prototype of the DGCP. The evaluation of the DGCP prototype indicated that the design requirements were fulfilled. Then, the evaluation of the final, workable DGCP showed that the stakeholders were satisfied with the system. Further, the evaluation also pointed out that the human values derived from the VSD approach were reflected in the DGCP. Contributions: The dissertation advances e-government design knowledge for sustainability. DGCP is an instantiation of an e-government tool for facilitating collaboration on environmental sustainability issues in Sri Lanka. The design knowledge is produced by incorporating human values and translating them into de

This paper investigates the real-time prediction of nitrogen oxides (NOx) emission by using around 17000 samples involved in a collection of three-day real data from a waste incineration power plant. To disclose the relationship between the ammonia (NH3) ejection and NOx emission, we choose the NOx reduction from inlet to outlet rather than the NOx concentration monitored by continuous emission monitoring system (CEMS). A hybrid procedure is developed to select appropriate features from the large and unsynchronized data, with which we establish a model based on the gradient boosting decision tree (GBDT) for the prediction. Computational experiments demonstrate that, with root mean square error (RMSE) values being 1.851 and 3.593 for training and test data, respectively, GBDT outperforms its two popular counterparts, supporting vector regression (SVR) and long short-term memory (LSTM). Shapley additive explanations (SHAP) is also conducted for analysis.

In public use of a large display, it is a usual phenomenon that multiple users individually participate in respective tasks on a common device. Previous studies have categorized such activity as independent interaction that involves little group engagement. However, by investigating how users approach, participate in, and interact with large displays, we found that parallel use is affected by group factors such as group size and between-user relationship. To gain a thorough understanding of individual and group behaviors, as well as parallel interaction task performance, one 70-inch display-based information searching task and experiment was conducted, in which a mobile eye movement tracking headset and a motion sensing RGB-depth sensor were simultaneously applied. The results showed that (1) a larger group size had a negative influence on the group users’ concentration on the task, perceived usability, and user experience; (2) a close relationship between users contributed to occasional collaborations, which was found to improve the users’ task completion time efficiency and their satisfaction on the large display user experience. This study proves that an integrated application of eye movement tracking and motion sensing is capable of understanding individual and group users’ behaviors simultaneously, and thus is a valid and reliable scheme in monitoring public activities that can be widely used in public large display systems.

Connected devices have become an integral part of modern homes and household devices, such as vacuum cleaners and refrigerators, are now often connected to networks. This connectivity introduces an entry point for cyber attackers. The plethora of successful cyber attacks against household IoT indicates that the security of these devices, or the security of applications related to these devices, is often lacking. Existing penetration testing studies usually focus on individual devices, and recent studies often mention the need for more extensive vulnerability assessments. Therefore, this study investigates the cyber security of devices commonly located in connected homes. Systematic penetration tests were conducted on 22 devices in five categories related to connected homes: smart door locks, smart cameras, smart car adapters/garages, smart appliances, and miscellaneous smart home devices. In total, 17 vulnerabilities were discovered and published as new CVEs. Some CVEs received critical severity rankings from the National Vulnerability Database (NVD), reaching 9.8/10. The devices are already being sold and used worldwide, and the discovered vulnerabilities could lead to severe consequences for residents, such as an attacker gaining physical access to the house. In addition to the published CVEs, 52 weaknesses were discovered that could potentially lead to new CVEs in the future. To our knowledge, this is the most comprehensive study on penetration testing of connected household products.

Molecular structure property modeling is an increasingly important tool for predicting compounds with desired properties due to the expensive and resource-intensive nature and the problem of toxicity-related attrition in late phases during drug discovery and development. Lately, the interest for applying deep learning techniques has increased considerably. This investigation compares the traditional physico-chemical descriptor and machine learning-based approaches through autoencoder generated descriptors to two different descriptor-free, Simplified Molecular Input Line Entry System (SMILES) based, deep learning architectures of Bidirectional Encoder Representations from Transformers (BERT) type using the Mondrian aggregated conformal prediction method as overarching framework. The results show for the binary CATMoS non-toxic and very-toxic datasets that for the former, almost equally balanced, dataset all methods perform equally well while for the latter dataset, with an 11-fold difference between the two classes, the MolBERT model based on a large pre-trained network performs somewhat better compared to the rest with high efficiency for both classes (0.93–0.94) as well as high values for sensitivity, specificity and balanced accuracy (0.86–0.87). The descriptor-free, SMILES-based, deep learning BERT architectures seem capable of producing well-balanced predictive models with defined applicability domains. This work also demonstrates that the class imbalance problem is gracefully handled through the use of Mondrian conformal prediction without the use of over- and/or under-sampling, weighting of classes or cost-sensitive methods.

Purpose - The purpose of this commentary is to comment on Fischer's et al. (2022) Design/methodology/approach - This commentary responds to Fischer's et al. (2022) call on envisioning alternate conceptualizations of learning for the digital era. In doing so, the author argues for reconsidering learning in its socio-material condition, situated and made of a web of social and technological relations. In this context, the author takes a relational lens on learning to interrogate taken-for-granted views of (1) personalizing data increasingly used for student learning, (2) emerging educational infrastructures for higher education and (3) the student–teacher relationship mediated by data and algorithms. Findings - In this commentary, the author suggested unpacking assumptions about learning that get reflected in the design and discourses about socio-technical arrangements and transformations in education. Taking the example of personalized learning, the author has illustrated a relational mode of thinking that leads the author to argue that, renewed definitions of learning must be discussed multidimensionally and, most importantly, situated in the material world that learning is already part of. Research limitations/implications - Following Fischer et al. (2022, this issue), the author agrees that the focus should be on finding “new ways of organizing learning by exploring opportunities for radically new conceptualizations and practices.” In order to do that it is of utmost importance to problematize the social and material conditions that actively configure learning today and infrastructure tomorrow's learning. Hopefully, these observations will entice others to discuss further the educational transformations at stake in the age of datafication and algorithmic decision-making. Originality/value - The author argues for reconsidering learning in its socio-material condition, which is situated and made of a web of social and technological relations. In this context, the au

The concept of the Internet of Medical Things brings a promising option to utilize various electronic health records stored in different medical devices and servers to create practical but secure clinical decision support systems. To achieve such a system, we need to focus on several aspects, most notably the usability aspect of deploying it using low-end devices. This study introduces one such application, namely FedSepsis, for the early detection of sepsis using electronic health records. We incorporate several cutting-edge deep learning techniques for the prediction and natural-language processing tasks. We also explore the multimodality aspect for the better use of electronic health records. A secure distributed machine learning mechanism is essential to building such a practical internet of medical things application. To address this, we analyze two federated learning techniques. Moreover, we use two different kinds of low-computational edge devices, namely Raspberry Pi and Jetson Nano, to address the challenges of using such a system in a practical setting and report the comparisons. We report several critical system-level information about the devices, namely CPU utilization, disk utilization, process CPU threads in use, process memory in use (non-swap), process memory available (non-swap), system memory utilization, temperature, and network traffic. We publish the prediction results with the evaluation metrics area under the receiver operating characteristic curve, the area under the precision–recall curve, and the earliness to predict sepsis in hours. Our results show that the performance is satisfactory, and with a moderate amount of devices, the federated learning setting results are similar to the single server-centric setting. Multimodality provides the best results compared to any single modality in the input features obtained from the electronic health records. Generative adversarial neural networks provide a clear superiority in handling the sparsity

EyeCompass is a novel free-eye drawing system enabling high-fidelity and efficient free-eye drawing through unimodal gaze control, addressing the bottlenecks of gaze-control drawing. EyeCompass helps people to draw using only their eyes, which is of value to people with motor disabilities. Currently, there is no effective gaze-control drawing application due to multiple challenges including involuntary eye movements, conflicts between visuomotor transformation and ocular observation, gaze trajectory control, and inherent eye-tracking errors. EyeCompass addresses this using two initial gaze-control drawing mechanisms: brush damping dynamics and the gaze-oriented method. The user experiments compare the existing gaze-control drawing method and EyeCompass, showing significant improvements in the drawing performance of the mechanisms concerned. The field study conducted with motor-disabled people produced various creative graphics and indicates good usability of the system. Our studies indicate that EyeCompass is a high-fidelity, accurate, feasible free-eye drawing method for creating artistic works via unimodal gaze control.

Noisy gradient algorithms have emerged as one of the most popular algorithms for distributed optimization with massive data. Choosing proper step-size schedules is an important task to tune in the algorithms for good performance. For the algorithms to attain fast convergence and high accuracy, it is intuitive to use large step-sizes in the initial iterations when the gradient noise is typically small compared to the algorithm-steps, and reduce the step-sizes as the algorithm progresses. This intuition has been confirmed in theory and practice for stochastic gradient descent. However, similar results are lacking for other methods using approximate gradients. This paper shows that the diminishing step-size strategies can indeed be applied for a broad class of noisy gradient algorithms. Our analysis framework is based on two classes of systems that characterize the impact of the step-sizes on the convergence performance of many algorithms. Our results show that such step-size schedules enable these algorithms to enjoy the optimal rate. We exemplify our results on stochastic compression algorithms. Our experiments validate fast convergence of these algorithms with the step decay schedules.

Rapid technological advancements have paved the way for better support for human health and well-being, and ICT-supported occupational stress management is one such growing area. However, this field lacks studies that can provide pragmatic solutions for occupational stress management through the design of artefacts for specific occupational groups, especially in developing countries. Previous studies of ICT-supported interventions for occupational stress management have shown low engagement and adherence, which have reduced their potential benefits. The aim of this thesis is therefore to conceptualise, design and evaluate an ICT-supported intervention for occupational stress management that can support user engagement and adherence among Sri Lankan software employees. The software industry in Sri Lanka is a fast-growing sector that is prone to producing stressed employees due to the demanding nature of the work. Design science research methodology was used as a guide for the research activities in this study. The thesis consists of eight interrelated studies in which a diversified range of research strategies were applied. Data were captured from three stakeholder categories: software employees, human resource managers in Colombo-based software companies, and stress experts from private, non-governmental and governmental organisations. The data collection methods used in the studies varied, and included literature reviews, questionnaires, interviews, and focus group discussions. Inductive thematic analysis was applied to analyse the qualitative data, and descriptive statistics were used for the quantitative data. Based on these studies, the proposed artefact was conceptualised as a self-help intervention that could be used with little guidance to promote health and well-being. The artefact incorporates multiple functional modules that provide both a ‘personal space’ and a ‘collaborative space’. The personal space provides users with resources and tools to manage str

Background Communicable diseases pose a severe threat to public health and economic growth. The traditional methods that are used for public health surveillance, however, involve many drawbacks, such as being labor intensive to operate and resulting in a lag between data collection and reporting. To effectively address the limitations of these traditional methods and to mitigate the adverse effects of these diseases, a proactive and real-time public health surveillance system is needed. Previous studies have indicated the usefulness of performing text mining on social media. Objective To conduct a systematic review of the literature that used textual content published to social media for the purpose of the surveillance and prediction of communicable diseases. Methodology Broad search queries were formulated and performed in four databases. Both journal articles and conference materials were included. The quality of the studies, operationalized as reliability and validity, was assessed. This qualitative systematic review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Results Twenty-three publications were included in this systematic review. All studies reported positive results for using textual social media content to surveille communicable diseases. Most studies used Twitter as a source for these data. Influenza was studied most frequently, while other communicable diseases received far less attention. Journal articles had a higher quality (reliability and validity) than conference papers. However, studies often failed to provide important information about procedures and implementation. Conclusion Text mining of health-related content published on social media can serve as a novel and powerful tool for the automated, real-time, and remote monitoring of public health and for the surveillance and prediction of communicable diseases in particular. This tool can address limitations related to traditional surve

We demonstrate the construction of robust counterfactual explanations for support vector machines (SVM), where the privacy mechanism that publicly releases the classifier guarantees differential privacy. Privacy preservation is essential when dealing with sensitive data, such as in applications within the health domain. In addition, providing explanations for machine learning predictions is an important requirement within so-called high risk applications, as referred to in the EU AI Act. Thus, the innovative aspects of this work correspond to studying the interaction between three desired aspects: accuracy, privacy, and explainability. The SVM classification accuracy is affected by the privacy mechanism through the introduced perturbations in the classifier weights. Consequently, we need to consider a trade-off between accuracy and privacy. In addition, counterfactual explanations, which quantify the smallest changes to selected data instances in order to change their classification, may become not credible when we have data privacy guarantees. Hence, robustness for counterfactual explanations is needed in order to create confidence about the credibility of the explanations. Our demonstrator provides an interactive environment to show the interplay between the considered aspects of accuracy, privacy, and explainability.

A major issue in the digital age is how to safeguard the security of the massive amount of data being stored, processed and transferred through digital channels. Organisational communication research into data security shows that people are the weakest link. Non-compliant behaviour can lead to security breaches. Existing studies have focused above all on how employees can be motivated to comply with data security procedures. However, focusing on desirable behaviour does not explain why people often ignore security regulations. In addition, existing research tends to focus primarily on self-reported attitudes and perceptions, which can give a distorted impression of people's actual behaviour. In this paper, we address these issues by combining individual, habitual and situational factors to explain non-compliant behaviour in a vignette study, using SmartPLS to analyse survey data from 651 subjects in a large Dutch government organisation. The results indicate that bad habits play a significant role in non-compliant behaviour. This behaviour is fuelled by situational factors like time pressure, while a lack of self-efficacy also increases non-compliant behaviour. Based on these results, a communication strategy that addresses bad habits in a situational context may provide an alternative way to improve people's compliant behaviour.

Sleep takes one-third of our life and is of increasing technological interest. Yet development has been focused on the traditional notion of 8 hours-per-night schedules of good sleep' - overlooking the varied nature of people's sleeping and waking practices. Based on the experiences of Shift Workers, Gamers, and Polyphasic Sleepers, this paper exposes the sleep practices of extreme sleepers for the design of technology. This cohort follows sleep patterns outwith those encoded in current technology for work, will and well-being. Through interviews and design workshops, we present three themes emerging from these perspectives on sleep. These themes revolve around the physiological, felt, and social timings of sleep. We explore a design space around these themes through eight different critical design exemplars. In conclusion, this work encourages the design of sleep technology focusing on users' agency over their own life-rhythms rather than pre-built concepts of good' sleep.

Digital forensic investigations require that file metadata are interpreted correctly. In this paper we focus on the timestamps of the exFAT file system. How these timestamps are written may depend on the implementation of the file system. We have performed experiments using Windows, MacOS and Linux to examine whether the respective file system drivers for exFAT use timestamps in the same manner, and whether they take the directory entry UTCOffset fields into account. We have also studied whether the forensic tools: Autopsy, X-Ways Forensics, EnCase Examiner, and FTK Imager interpret the timestamps consistently. The results show that there are substantial inconsistencies both in the file system implementations and in how forensic tools handle these inconsistencies. For the unwary forensic examiner, there is a clear risk of interpreting timestamps incorrectly by a substantial margin. We conclude that timestamp interpretation during criminal investigations should not be based on the assumption that the file system specifications are followed flawlessly by the file system driver developers or necessarily interpreted and displayed correctly by the digital forensic tools.

This work-in-progress presents the design of an XR prototype for the purpose of educating basic cybersecurity concepts. We have designed an experimental virtual reality cyberspace to visualise data traffic over network, enabling the user to interact with VR representations of data packets. Our objective was to help the user better conceptualise abstract cybersecurity topics such as encryption and decryption, firewall and malicious data. Additionally, to better stimuli the sense of immersion we have used Peltier thermoelectric modules and Arduino Uno to experiment with multisensory XR. Furthermore, we reflect on early evaluation of this experimental prototype and present potential paths for future improvements.

This paper explores the design space of collaborative video editing through a series of design workshops with video editors. Collaborative video editing can be supported by adding awareness features or other well-known collaborative features found in existing software and introducing new features designed specifically for video editing software. The paper identifies different design concepts that illustrate how such collaborative features can be included in non-linear video editing software and discusses the challenges of introducing such features. Some design concepts are explicitly inspired by existing collaborative tools. However, we suggest that introducing such features might not be straightforward. In other cases, alternative abstract representations of time-based media might be necessary to support collaborative video editing.

Despite the growing concerns related to environmental disruptions, such as forest fires and floods, our interaction with nature has become increasingly distant. The belief is that this is due to human attunement to quick changes over short periods, while natural changes seem slow-paced over the long term. We argue that rather than the problem lying with the pace of time, humans should further attune to the various temporalities within nature. To unpack this topic, we follow a two-fold process: Research through Design and Autobiographical Design. The resulting prototype takes environmental sensor data and visualises the changes happening over a short time. We complement this IT-supported process with think-aloud walks, a practice that is embodied and devoid of technology, for a plural perspective on time. We offer insights into how noticing the various temporalities in nature was enabled, finally arguing for designs that support progressive adjustments that sustain human-nonhuman mutualism.

Interacting with long lists of instructions or ingredients continues to be a challenge for conversational interaction. In this paper, we conducted a user study to experiment with the use of 'cued-gaze'-waiting for the user's visual attention-to manage the delivery of instructions with a voice agent. In a Wizard-of-Oz setting, 12 participants were instructed to build a simple Lego tower by a conversational agent and were able to advance in the list using either speech interaction, or gaze interaction. The increasing use of speech agents in real-world cause users to encounter failures in interactions, so in this task the agent was designed to fail when providing the list of instruction to explore how the participants proceeded to recover from common failures. This showed that, for this use case, cross-modality repair was more effective than reformulation of speech.

This paper presents a study of a group of university students using generative machine learning to translate from natural language to computer code. The study explores how the use of the AI tool can be understood in terms of co-creation, focusing on the one hand on how the tool may serve as a resource for understanding and learning, and on the other hand how the tool affects the creative processes. Findings show how the participants search for a 'correct' syntax in their instructions to the machine learning tool, and how the inconsistent and erroneous behavior can work as a way to generate clues and inspiration for generating creative expressions. The notion of friction is used to describe how systems like this can serve to both lower thresholds for programming, and also interfere with the creative processes, encouraging reflection and exploration of alternative solutions.

A survey on trust management in the Social Internet of Things (SIoT) is provided, beginning with a discussion of SIoT architectures and relationships. Using a variety of publication databases, we describe efforts that focus on various trust management aspects of SIoT. Trust management models comprise three themes: trust computation, aggregation, and updates. Our study presents a detailed discussion of all three steps. Trust computation and trust aggregation depend upon Trust Attributes (TAs) for the calculation of local and global trust values. Our paper discusses many strategies for aggregating trust, but “Weighted Sum” is the most frequently used in the relevant studies. Our paper addresses trust computation and aggregation scenarios. Our work classifies research by TAs (Social Trust, Quality of Service). We’ve categorized the research (reputation-based, recommendation-based, knowledge-based) depending on the types of feedback/opinions used to calculate trust values (global feedback/opinion, feedback from a friend, trustor’s own opinion considering the trustee’s information). Our work classifies studies (policy-based, prediction-based, weighted sum-based/weighted linear combination-based) by trust computation/aggregation approach. Two trust-update schemes are discussed: time-driven and event-driven schemes, while most trust management models utilize an event-driven scheme. Both trust computation and aggregation need propagating trust values in a centralized, decentralized, or semi-centralized way. Our study covers classifying research by trust updates and propagation techniques. Trust models should provide resiliency to SIoT attacks. This analysis classifies SIoT attacks as collaborative or individual. We also discuss scenarios depicted in the relevant studies to incorporate resistance against trust-related attacks in SIoT. Studies suggest context-based or context-free trust management strategies. Our study categorizes studies based on context-based or context-fre

This paper presents Fractured Objects for the design of virtual and mixed-reality experiences. Drawing on the qualitative analysis of three weeks of artistic activities within a residency program, we present six types of Fractured Objects that were used in sketching a mixed-reality performance. Building on these Fractured Objects, as they were articulated by the artists, we present speculative designs for their use in scenarios inspired by research within the IMX community. In discussion, we look to expand the concept of Fractured Objects by relating it to other design concepts such as Seamful Design and Wabi-Sabi, and explore the relationship to the temporality of interaction. We introduce Kintsugi VR with Fractured Objects, drawing on the concept of golden repair' in which the act of reconnecting fractured parts improves the resulting whole object.

The semiconductor industry is well known for its high resource consumption due to its clean room usage and power-intensive manufacturing technology. This is a significant financial burden for semiconductor fabrication plant owners who wish to minimize the cost needed to cover the electricity and gas loads for their daily operations. From a historical perspective, those electricity and gas cost reductions have been enabled by more energy-efficient manufacturing technologies. Due to the complexity of factors influencing the energy system of a semiconductor plant, other aspects have often been neglected and also not been studied intensively in the academic literature so far. One possibility for reducing energy costs is the integration of a microgrid as on-site distributed energy resources (DER) may offer a less cost-intensive way to supply the energy needed. Therefore, this paper seeks to answer the following research question: How much expected cost can be saved and risk mitigated at a semiconductor fabrication plant via DER? More precisely, in this paper, a microgrid at a semiconductor fabrication plant that has installed DER with combined heat and power (CHP) applications and demand response (DR) is used to simulate the expected annual energy costs under given uncertain electricity and gas prices at different installed capacity levels. A simulation-based approach is used to evaluate several DER capacity sizes. The results of the simulation show that installations of DER at a semiconductor microgrid can dominate the base case of doing nothing, considering the expected minimized cost, the expected CO2 emissions, and conditional value-at-risk (CVaR). In fact, DER can reduce the expected annual energy bill by up to 6%, whereas annual expected CO2 emissions can even be reduced by up to 22%. Moreover, with DER, a microgrid’s risk is reduced as it can react to market conditions and local demand. Additionally, the true potential of microgrid cost savings can only be enabled, Die Halbleiterindustrie ist aufgrund der Nutzung von Reinräumen und der energieintensiven Fertigungstechnologie bekannt für ihren hohen Ressourcenverbrauch. Dies stellt einen erheblichen finanziellen Aufwand für die Eigentümer von Halbleiterproduktionsanlagen dar, die die Kosten für die Deckung der Energie- und Gaslast ihres täglichen Betriebes minimieren möchten. Aus historischer Sicht wurden diese Energie- und Gaskostensenkungen durch energieeffizientere Fertigungstechnologien ermöglicht. Aufgrund der Komplexität der Faktoren, die das Energiesystem eines Halbleiterwerks beeinflussen, wurden andere Aspekte oft vernachlässigt und auch in der wissenschaftlichen Literatur bisher nicht intensiv untersucht. Eine mögliche Alternative zur Minimierung der Energiekosten ist die Integration eines Mikronetzes, da dezentrale Energieressourcen (DER) vor Ort eine weniger kostenintensive Möglichkeit zur Bereitstellung der benötigten Energie bieten können. In diesem Beitrag wird daher versucht, die folgende Frage zu beantworten: Wie hoch sind die erwarteten Kosteneinsparungen und die Risikominderung in einer Halbleiterfabrik durch eine Investition in DER? Genauer gesagt, wird in diesem Beitrag ein Mikronetz in einem Halbleiterwerk, das DER mit Kraft-Wärme-Kopplung (KWK) und Laststeuerung (LS) installiert hat, verwendet, um die erwarteten jährlichen Energiekosten unter gegebenen unsicheren Strom- und Gaspreisen bei verschiedenen installierten Kapazitätsniveaus zu simulieren. Ein simulationsbasierter Ansatz wird verwendet, um verschiedene DER-Kapazitätsinvestitionen zu bewerten. Die Simulationsergebnisse zeigen, dass DER-Installationen in einem Halbleiter-Mikronetz den Basisfall des Nichtstuns dominieren können, wenn man die erwarteten minimierten Kosten, die erwarteten CO2-Emissionen und den Conditional Value-at-Risk (CVaR) berücksichtigt. In der Tat kann DER die erwarteten jährlichen Energiekosten um bis zu 6 % senken, während die erwarteten jährlichen CO2-Emissionen sogar um bis

This article proposes a new workflow for Culture Heritage (CH) Collection Data in which distributed meta-learning with context-based fusion is required. The multidisciplinary research outlines a new paradigm for collaborative creation of a context- based data fusion CH workflow. Whose criterion restricts the search space to the cultural sector to include multiple other CH object types, texts and predominantly analyzes visualizations of CH object metadata. The approach promotes user-driven content creation and offsets economic models, thereby rewarding creators and performers. In response to these challenges, we propose a framework for bringing about massive and real-time presence and awareness on the Internet through an Internet of Things infrastructure to connect objects, context and intangible assets. We enable an online virtual world that incorporates proximity of pervasive information, objects, people, processes, data and places. Finally, we investigate some new ways to achieve immersive experiences via distributed meta-learning computing and point out the necessity to do more with regard to collaborative interaction creation.

Computer assisted coding (CAC) of clinical text into standardized classifications such as ICD-10 is an important challenge. For frequently used ICD-10 codes, deep learning approaches have been quite successful. For rare codes, however, the problem is still outstanding. To improve performance for rare codes, a pipeline is proposed that takes advantage of the ICD-10 code hierarchy to combine semantic capabilities of deep learning and the flexibility of fuzzy logic. The data used are discharge summaries in Swedish in the medical speciality of gastrointestinal diseases. Using our pipeline, fuzzy matching computation time is reduced and accuracy of the top 10 hits of the rare codes is also improved. While the method is promising, further work is required before the pipeline can be part of a usable prototype. Code repository: https://github.com/icd-coding/zeroshot.

Large, varied, and time-evolving data sources can be observed across many domains and present a unique challenge for classification problems, in which traditional machine learning approaches must be adapted to accommodate for the complex nature of such data. Across most domains, there is also a need for machine learning models that are both well-performing and interpretable, to help provide explanations of a model's decisions that stakeholders can trust and take appropriate actions with. In the medical domain, complex Electronic Health Record (EHR) data consists of longitudinal records of patient histories spanning structured and unstructured data types. Exploiting such complex medical data is vital as a means of gaining useful medical insights and predictions, and where establishing stakeholder trust through useful explanations is critical. This thesis has focused on producing state-of-the-art classification methods for exploiting the heterogeneity and temporality of complex data; secondly, on developing novel interpretability methods to aid in the understanding of model predictions from such complex data; and finally on ensuring the medical applicability of the developed methods and other novel methods particularly for the medical problem of adverse drug event (ADE) prediction. In the first part of this thesis, several state-of-the-art classification frameworks for exploiting complex medical data are outlined, with their utility demonstrated through comparative empirical evaluations to competing framework approaches. In the second part of this thesis, novel interpretability methods are developed and demonstrated for their applicability across domains. In the third part of this thesis, the applicability of interpretability and explanability methods for complex medical data are investigated, refined, and assessed for validity in connection to the use-case of ADE prediction. Main contributions of this thesis include: two novel classification frameworks, including SM

We propose a context-sensitive deliberation framework where the decision context does not deliver an action straight away, but where rather the decision context and agent characteristics influence the type of deliberation and type of information evaluated which will affect the final decision. The framework is based on the Contextual Action Framework for Computational Agents (CAFCA). Our framework also tailors the deliberation type used to the decision context the agent finds itself in, starting from the least cognitive taxing deliberation types unless the context requires more complex deliberation types. As a proof-of-concept the paper shows how context and information relevance can be used to conceptually expand the deliberation system of an agent.

This article explains the new features of the Excite-O-Meter, an open-source tool that enables the collection of bodily data, real-time feature extraction, and post-session data visualization in any custom VR environment developed in Unity. Besides analyzing heart activity, the tool supports now multidimensional time series to study motion trajectories in VR. The paper presents the main functionalities and discusses the relevance of the tool for behavioral and psychophysiological research.

In public procurement, byers typically try to resolve the problem of not knowing the tenders before stating their preferences through the use of price relative models or comparison prices and different procurements use slightly different spreadsheet models for tender evaluation. This renders it difficult to assess the quality attributes in a way enabling for evaluating and comparing price-quality properties of real-life tender evaluations. However, tender evaluation is an instance of multi-attribute decision-making, and for each tender evaluation instance there is a corresponding multi-attribute decision model enabling for a more systematic assessment of tender evaluations. In this paper, we investigate how robust the selection of tenders in IT procurement are towards changes in the assessment of quality and to the presence of quality attributes in the tender evaluation model. Conclusions include that "decisive"attributes cannot be expected to exist, and that score based tender evaluation model with precise estimates of quality levels put strong demands on the capabilities of the buyer.