Search

Predictive student models are increasingly used in learning environments due to their ability to enhance educational outcomes and support stakeholders in making informed decisions. However, predictive models can be biased and produce unfair outcomes, leading to potential discrimination against certain individuals and harmful long-term implications. This has prompted research on fairness metrics meant to capture and quantify such biases. Nonetheless, current metrics primarily focus on predictive performance comparisons between groups, without considering the behavior of the models or the severity of the biases in the outcomes. To address this gap, we proposed a novel metric in a previous work (Verger et al., 2023) named "Model Absolute Density Distance" (MADD), measuring algorithmic unfairness as the difference of the probability distributions of the model's outcomes. In this paper, we extended our previous work with two major additions. Firstly, we provided theoretical and practical considerations on a hyperparameter of MADD, named "bandwidth," useful for optimal measurement of fairness with this metric. Secondly, we demonstrated how MADD can be used not only to measure unfairness but also to mitigate it through postprocessing of the model's outcomes while preserving its accuracy. We experimented with our approach on the same task of predicting student success in online courses as our previous work, and obtained successful results. To facilitate replication and future usages of MADD in different contexts, we developed an open-source Python package called maddlib (https://pypi.org/project/maddlib/). Altogether, our work contributes to advancing the research on fair student models in education.

In this paper, we present an extended evaluation of a course recommender system designed to support students who struggle in the first semesters of their studies and are at risk of dropping out. The system, which was developed in earlier work using a student-centered design, is based on the explainable k-nearest neighbor algorithm and recommends a set of courses that have been passed by the majority of successful neighbors, that is, students who graduated from the study program. In terms of the number of recommended courses, we found a discrepancy between the number of courses that struggling students are recommended to take and the actual number of courses they take. This indicates that there may be an alternative path that these students could consider. However, the recommended courses align well with the courses taken by students who successfully graduated. This suggests that even students who are performing well could still benefit from the course recommender system designed for at-risk students. In the present work, we investigate a second type of success--a specific minimum number of courses passed--and compare the results with our first approach from previous work. With the second type, the information about success might be already available after one semester instead of after graduation which allows faster growth of the database and faster response to curricular changes. The evaluation of three different study programs in terms of dropout risk reduction and recommendation quality suggests that course recommendations based on students passing at least three courses in the following semester can be an alternative to guide students on a successful path.

In order to improve the effect of intelligent automatic test paper composition, this paper combines the hybrid fuzzy clustering algorithm to study the computer automatic test paper composition algorithm. In this paper, a computer automatic test paper composition system based on hybrid fuzzy clustering algorithm is constructed. Moreover, the hybrid fuzzy clustering method used in this paper is used as the basic algorithm of the system, and the algorithm is improved according to the actual needs of intelligent paper composition. In addition, this paper uses an intelligent algorithm to input the relevant constraint parameters and combines the original parameters to select the most suitable test questions from the database and combine them into test papers. Finally, this paper constructs the system structure based on the requirements of intelligent test paper composition. The experimental research shows that the computer automatic test paper composition system based on the hybrid fuzzy clustering algorithm proposed in this paper has a good test paper composition function, which can effectively promote the progress of the intelligent examination mode in colleges and universities. [ABSTRACT FROM AUTHOR]

This special issue contains six revised and extended versions of tool papers that appeared in the proceedings of TACAS 2021, the 27th International Conference on Tools and Algorithms for the Construction and Analysis of Systems. The issue is dedicated to the realization of algorithms in tools and the studies of the application of these tools for analysing hard- and software systems. [ABSTRACT FROM AUTHOR]

This research was triggered by the identified need in literature for large-scale studies about the kinds of designs that teachers create for mobile learning (m-learning). These studies require analyses of large datasets of learning designs. The common approach followed by researchers when analyzing designs has been to manually classify them following high-level pedagogically guided coding strategies, which demands extensive work. Therefore, the first goal of this paper is to explore the use of supervised machine learning (SML) to automatically classify the textual content of m-learning designs using pedagogically relevant classifications, such as the cognitive level demanded by students to carry out specific designed tasks, the phases of inquiry learning represented in the designs, or the role that the situated environment has in the designs. Because not all SML models are transparent, but researchers often need to understand their behaviour, the second goal of this paper is to consider the trade-off between models' performance and interpretability in the context of design analytics for m-learning. To achieve these goals, we compiled a dataset of designs deployed using two tools, Avastusrada and Smartzoos. With this dataset, we trained and compared different models and feature extraction techniques. We further optimized and compared the best performing and most interpretable algorithms (EstBERT and Logistic Regression) to consider the second goal with an illustrative case. We found that SML can reliably classify designs with accuracy > 0.86 and Cohen's kappa > 0.69.

The continual evolution of employment opportunities in the present industrial era has raised the need for career-long expert advice. Similar to other fields, thankfully technology has come to our rescue in the area of career guidance also. This paper presents a systematic review of Expert Systems (ES) developed for career guidance, course selection and evaluation of students in the past ten years. The popular research databases Google Scholar and Science Direct were used for obtaining the relevant research papers through broad keywords. The keywords were refined to identify the articles related to rule-based, case-based and fuzzy logic-based ES used for career guidance. A total of twenty-five peer-reviewed relevant articles with full-text available online was selected for the final study. In order to avoid duplicity, technical reports and unreferenced literature were excluded. The review identifies the relatively high weight given by the researchers to rule-based systems owing to their simplicity and broad applicability. However, the relative merits and demerits of rule-based, case-based and fuzzy logic-based ES are highly dependent on the field of application. Nevertheless, ES find wide applications in the area of career guidance and have the potential to enhance the career guidance accessibility of the most remote students.

To link to full-text access for this article, visit this link: http://dx.doi.org/10.1016/j.ejor.2007.10.019 Byline: Kazuki Matsumoto (a), Hiroyoshi Miwa (b), Toshihide Ibaraki (b) Keywords: Scheduling; Corrugated paper; Complexity; Contiguous set; Integer programming Abstract: Corrugated paper is produced by gluing three types of papers of the same breadth. Given a set of orders, we first assign each order to one of the standard breadths, and then sequence those assigned to each standard breadth so that they are continuously manufactured from the three rolls of the specified standard breadth equipped in the machine called corrugator. Here we are asked to achieve multi-goals of minimizing total length of roll papers, total loss of papers caused by the differences between standard breadths and real breadths of the orders, and the number of machine stops needed during production. We use integer programming to assign orders to standard breadths, and then develop a special purpose algorithm to sequence the orders assigned to each standard breadth. This is a first attempt to handle scheduling problems of the corrugator machine. Author Affiliation: (a) Department of Applied Mathematics and Physics, Graduate School of Informatics, Kyoto University, Sakyo-ku, Kyoto 606-8501, Japan (b) Department of Informatics, School of Science and Technology, Kwansei Gakuin University, 2-1 Gakuen Sanda, Hyogo 669-1337, Japan Article History: Received 24 March 2007; Accepted 5 October 2007

Artificial intelligence (AI) has the potential to positively impact library and archives collections and services--enhancing reference, instruction, metadata creation, recommendations, and more. However, AI also has ethical implications. This paper presents an extensive literature and review analysis that examines AI projects implemented in library and archives settings, asking the following research questions: RQ1: How is artificial intelligence being used in libraries and archives practice? RQ2: What ethical concerns are being identified and addressed during AI implementation in libraries and archives? The results of this literature review show that AI implementation is growing in libraries and archives and that practitioners are using AI for increasingly varied purposes. We found that AI implementation was most common in large, academic libraries. Materials used in AI projects usually involved digitized and born digital text and images, though materials also ranged to include web archives, electronic theses and dissertations (ETDs), and maps. AI was most often used for metadata extraction and reference and research services. Just over half of the papers included in the literature review mentioned ethics or values related issues in their discussions of AI implementation in libraries and archives, and only one-third of all resources discussed ethical issues beyond technical issues of accuracy and human-in-the-loop. Case studies relating to AI in libraries and archives are on the rise, and we expect subsequent discussions of relevant ethics and values to follow suit, particularly growing in the areas of cost considerations, transparency, reliability, policy and guidelines, bias, social justice, user communities, privacy, consent, accessibility, and access. As AI comes into more common usage, it will benefit the library and archives professions to not only consider ethics when implementing local projects, but to publicly discuss these ethical considerations in shared documentation and publications. [ABSTRACT FROM AUTHOR]

The folded paper-size illusion is as easy to demonstrate as it is powerful in generating insights into perceptual processing: First take two A4 sheets of paper, one original sized, another halved by folding, then compare them in terms of area size by centering the halved sheet on the center of the original one! We perceive the larger sheet as far less than double (i.e., 100%) the size of the small one, typically only being about two thirds larger--this illusion is preserved by rotating the inner sheet and even by aligning it to one or two sides, but is dissolved by aligning both sheets to three sides, here documented by 88 participants' data. A potential explanation might be the general incapability of accurately comparing more than one geometrical dimension at once--in everyday life, we solve this perceptual-cognitive bottleneck by reducing the complexity of such a task via aligning parts with same lengths. [ABSTRACT FROM AUTHOR]

Compared to the conventional boost power factor correction (PFC) converter, a totem-pole bridgeless PFC has high efficiency because it does not have an input diode rectifier stage, but a current spike may occur when the polarity of the grid voltage changes. This paper proposes a digital control algorithm for bridgeless totem-pole PFC with a simple control structure based on the phase angle of grid voltage. The proposed algorithm has a PI-based double-loop control structure and performs DC-link voltage and input inductor current control. Rectifying switches operate based on the proposed rectification algorithm using phase angle information calculated through a single-phase phase-locked loop (PLL) to prevent current spikes. The feed-forward duty ratio value is calculated according to the polarity of the grid voltage and added to the double-loop controller to perform appropriate power factor control. The performance and feasibility of the proposed control algorithm are verified through a 3 kW hardware prototype. [ABSTRACT FROM AUTHOR]

Interpreting a seemingly simple function word like "or," "behind," or "more" can require logical, numerical, and relational reasoning. How are such words learned by children? Prior acquisition theories have often relied on positing a foundation of innate knowledge. Yet recent neural-network-based visual question answering models apparently can learn to use function words as part of answering questions about complex visual scenes. In this paper, we study what these models learn about function words, in the hope of better understanding how the meanings of these words can be learned by both models and children. We show that recurrent models trained on visually grounded language learn gradient semantics for function words requiring spatial and numerical reasoning. Furthermore, we find that these models can learn the meanings of logical connectives and and or without any prior knowledge of logical reasoning as well as early evidence that they are sensitive to alternative expressions when interpreting language. Finally, we show that word learning difficulty is dependent on the frequency of models' input. Our findings offer proof-of-concept evidence that it is possible to learn the nuanced interpretations of function words in a visually grounded context by using non-symbolic general statistical learning algorithms, without any prior knowledge of linguistic meaning.

In 3D building models, a large number of texture maps with different sizes increase the number of model data loading and drawing batches, which greatly reduces the drawing efficiency of the model. Therefore, this paper proposes a texture set mapping method based on vertex importance. Firstly, based on the 2D space boxing algorithm, the texture maps are merged and a series of Mipmap texture maps are generated, and then the vertex curvature, texture variability and location information of each vertex are calculated, normalized, and weighted to get the importance of each vertex, and then finally, different Mipmap-level textures are remapped according to the importance of the vertices. The experiment proves that the algorithm in this paper can reduce the amount of texture data on the one hand, and avoid the rendering pressure brought by the still large amount of data after merging on the other hand, so as to improve the rendering efficiency of the model. [ABSTRACT FROM AUTHOR]

The traditional method for transformer moisture diagnosis is to establish empirical equations between feature parameters extracted from frequency domain spectroscopy (FDS) and the transformer’s moisture content. However, the established empirical equation may not be applicable to a novel testing environment, resulting in an unreliable evaluation result. In this regard, it is acknowledged that FDS combined with machine learning is more suitable for estimating moisture content in a variety of test environments. Nonetheless, the accuracy of the estimation results obtained using the existing method is limited by the algorithm’s inability to generalize. To address this issue, we propose an AdaBoost algorithm-enhanced back-propagation neural network (BP_AdaBoost). This study creates a database by extracting feature parameters from the FDS that characterize the insulation states of the prepared samples. Then, using the BP_AdaBoost algorithm and the newly constructed database, the moisture estimation models are trained. Finally, the results of the estimation are discussed in terms of laboratory and field transformers. By comparing the proposed BP_AdaBoost algorithm to other intelligence algorithms, it is demonstrated that it not only performs better in generalization, but also maintains a high level of accuracy. [ABSTRACT FROM AUTHOR]

OBJECTIVE: To develop an automated algorithm to detect, quantify, and differentiate between tremor using pen-on-paper spirals. METHODS: Patients with essential tremor (n = 25), dystonic tremor (n = 25), Parkinson’s disease (n = 25), and healthy volunteers (HV, n = 25) drew free-hand spirals. The algorithm derived the mean deviation (MD) and tremor variability from scanned images. MD and tremor variability were compared with 1) the Bain and Findley scale, 2) the Fahn–Tolosa–Marin tremor rating scale (FTM–TRS), and 3) the peak power and total power of the accelerometer spectra. Inter and intra loop widths were computed to differentiate between the tremor. RESULTS: MD was higher in the tremor group (48.9±26.3) than in HV (26.4±5.3; p < 0.001). The cut-off value of 30.3 had 80.9% sensitivity and 76.0% specificity for the detection of the tremor [area under the curve: 0.83; 95% confidence index (CI): 0.75, 0.91, p < 0.001]. MD correlated with the Bain and Findley ratings (rho = 0.491, p = 0 < 0.001), FTM–TRS part B (rho = 0.260, p = 0.032) and accelerometric measures of postural tremor (total power, rho = 0.366, p < 0.001; peak power, rho = 0.402, p < 0.001). Minimum Detectable Change was 19.9%. Inter loop width distinguished Parkinson’s disease spirals from dystonic tremor (p < 0.001, 95% CI: 54.6, 211.1), essential tremor (p = 0.003, 95% CI: 28.5, 184.9), or HV (p = 0.036, 95% CI: -160.4, -3.9). CONCLUSION: The automated analysis of pen-on-paper spirals generated robust variables to quantify the tremor and putative variables to distinguish them from each other. SIGNIFICANCE: This technique maybe useful for epidemiological surveys and follow-up studies on tremor. [ABSTRACT FROM AUTHOR]

We study the quadratic cycle cover problem (QCCP), which aims to find a node-disjoint cycle cover in a directed graph with minimum interaction cost between successive arcs. We derive several semidefinite programming (SDP) relaxations and use facial reduction to make these strictly feasible. We investigate a nontrivial relationship between the transformation matrix used in the reduction and the structure of the graph, which is exploited in an efficient algorithm that constructs this matrix for any instance of the problem. To solve our relaxations, we propose an algorithm that incorporates an augmented Lagrangian method into a cutting-plane framework by utilizing Dykstra's projection algorithm. Our algorithm is suitable for solving SDP relaxations with a large number of cutting-planes. Computational results show that our SDP bounds and efficient cutting-plane algorithm outperform other QCCP bounding approaches from the literature. Finally, we provide several SDP-based upper bounding techniques, among which is a sequential Q-learning method that exploits a solution of our SDP relaxation within a reinforcement learning environment. Summary of Contribution: The quadratic cycle cover problem (QCCP) is the problem of finding a set of node-disjoint cycles covering all the nodes in a graph such that the total interaction cost between successive arcs is minimized. The QCCP has applications in many fields, among which are robotics, transportation, energy distribution networks, and automatic inspection. Besides this, the problem has a high theoretical relevance because of its close connection to the quadratic traveling salesman problem (QTSP). The QTSP has several applications, for example, in bioinformatics, and is considered to be among the most difficult combinatorial optimization problems nowadays. After removing the subtour elimination constraints, the QTSP boils down to the QCCP. Hence, an in-depth study of the QCCP also contributes to the construction of strong bounds for the QTSP. In this paper, we study the application of semidefinite programming (SDP) to obtain strong bounds for the QCCP. Our strongest SDP relaxation is very hard to solve by any SDP solver because of the large number of involved cutting-planes. Because of that, we propose a new approach in which an augmented Lagrangian method is incorporated into a cutting-plane framework by utilizing Dykstra's projection algorithm. We emphasize an efficient implementation of the method and perform an extensive computational study. This study shows that our method is able to handle a large number of cuts and that the resulting bounds are currently the best QCCP bounds in the literature. We also introduce several upper bounding techniques, among which is a distributed reinforcement learning algorithm that exploits our SDP relaxations. [ABSTRACT FROM AUTHOR]

This work describes the log-MAP (BCJR) algorithm implementation of a close to capacity dirty paper coding CODEC. The CODEC consists of eight deep pipeline processors. It decodes blocks of 975 bits in 26.9 ms using less than 9.7% of low-cost FPGA (and no DSP blocks). Two pipelines, for alpha and beta, calculate the values of gamma (of the BCJR) to reduce the storage requirements. The final log-likelihood ratio (LLR) is calculated together with alpha, reusing intermediate results. The number of bits used by the different signals of the processor is easily configurable. It was set to six bits to the channel measure signals and eight bits to log of probability signals like alpha, beta, and others. The CODEC clock was 100 MHz. The achieved bit rate is 36.2 Kbps per CODEC, but multiple CODECs can be fit into a single chip. The CODEC is 3.49 dB from the channel capacity. [ABSTRACT FROM AUTHOR]

Accurate characterization is an important issue in paper currency recognition system. This paper proposes a robust paper currency recognition method based on Hidden Markov Model (HMM). By employing HMM, the texture characteristics of paper currencies are modeled as a random process. The proposed algorithm can be used for distinguishing paper currency from different countries. A similarity measure has been used for the classification in the proposed algorithm. To evaluate the performance of the proposed algorithm, experiments have been conducted on more than 100 denominations from different countries. The results indicate 98% accuracy for recognition of paper currency. [Copyright &y& Elsevier]

Abstract: More than just being a substrate, paper can also provide evidence for the provenance of documents. An earlier paper described a method to compare paper structure, based on the Fourier power spectra of light transmission images. Good results were obtained by using the 2D correlation of images derived from the power spectra as a similarity score, but the method was very computationally intensive. Different comparison algorithms are evaluated in this paper, using information theoretical criteria. An angular invariant algorithm turned out to be as effective as the original one but 4 orders of magnitude faster, making the use of much larger databases possible. [Copyright &y& Elsevier]

In recent years, smart learning environments have become central to modern education and support students and instructors through tools based on prediction and recommendation models. These methods often use learning material metadata, such as the knowledge contained in an exercise which is usually labeled by domain experts and is costly and difficult to scale. It recognizes that automated labeling eases the workload on experts, as seen in previous studies using automatic classification algorithms for research papers and Japanese mathematical exercises. However, these studies didn't delve into fine-grained labeling. In addition to that, as the use of materials in the system becomes more widespread, paper materials are transformed into PDF formats, which can lead to incomplete extraction. However, there is less emphasis on labeling incomplete mathematical sentences to tackle this problem in the previous research. This study aims to achieve precise automated classification even from incomplete text inputs. To tackle these challenges, we propose a mathematical exercise labeling algorithm that can handle detailed labels, even for incomplete sentences, using word n-grams, compared to the state-of-the-art word embedding method. The results of the experiment show that mono-gram features with Random Forest models achieved the best performance with a macro F-measure of 92.50%, 61.28% for 24-class labeling and 297-class labeling tasks, respectively. The contribution of this research is showing that the proposed method based on traditional simple n-grams has the ability to find context-independent similarities in incomplete sentences and outperforms state-of-the-art word embedding methods in specific tasks like classifying short and incomplete texts.

In this paper, a key node mining algorithm of entropy-CRITIC combined weighted GRA-TOPSIS method is proposed, which is based on the network structure features. First, the method obtained multi-dimensional data of students' identities, seating relationships, social relationships, and so on to build a database. Then, the seating similarity among students was used to construct the in-class social networks and analyze the structural characteristics of them. Finally, the CRITIC and entropy weight method was introduced for obtaining the combined weight values and the GRA-TOPSIS multi-decision fusion algorithm to mine the key student nodes that have negative impact. The experiments showed that the algorithm of this paper can evaluate students objectively based on their classroom social networks, providing technical support for process-oriented comprehensive quality education evaluation.

A common divide-and-conquer approach for Bayesian computation with big data is to partition the data, perform local inference for each piece separately, and combine the results to obtain a global posterior approximation. While being conceptually and computationally appealing, this method involves the problematic need to also split the prior for the local inferences; these weakened priors may not provide enough regularization for each separate computation, thus eliminating one of the key advantages of Bayesian methods. To resolve this dilemma while still retaining the generalizability of the underlying local inference method, we apply the idea of expectation propagation (EP) as a framework for distributed Bayesian inference. The central idea is to iteratively update approximations to the local likelihoods given the state of the other approximations and the prior. The present paper has two roles: we review the steps that are needed to keep EP algorithms numerically stable, and we suggest a general approach, inspired by EP, for approaching data partitioning problems in a way that achieves the computational benefits of parallelism while allowing each local update to make use of relevant information from the other sites. In addition, we demonstrate how the method can be applied in a hierarchical context to make use of partitioning of both data and parameters. The paper describes a general algorithmic framework, rather than a specific algorithm, and presents an example implementation for it.

With the rapid development of information technology, scientific social networks (SSNs) have become the fastest and most convenient way for researchers to communicate with each other. Many published papers are shared via SSNs every day, resulting in the problem of information overload. How to appropriately recommend personalized and highly valuable papers for researchers is becoming more urgent. However, when recommending papers in SSNs, only a small amount of positive instances are available, leaving a vast amount of unlabelled data, in which negative instances and potential unseen positive instances are mixed together, which naturally belongs to One-Class Collaborative Filtering (OCCF) problem. Therefore, considering the extreme data imbalance and data sparsity of this OCCF problem, a hybrid approach of Social and Content aware One-class Recommendation of Papers in SSNs, termed SCORP, is proposed in this study. Unlike previous approaches recommended to address the OCCF problem, social information, which has been proved playing a significant role in performing recommendations in many domains, is applied in both the profiling of content-based filtering and the collaborative filtering to achieve superior recommendations. To verify the effectiveness of the proposed SCORP approach, a real-life dataset from CiteULike was employed. The experimental results demonstrate that the proposed approach is superior to all of the compared approaches, thus providing a more effective method for recommending papers in SSNs. [ABSTRACT FROM AUTHOR]

Copyright of Revista Catalana de Dret Públic is the property of Revista Catalana de Dret Public and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Purpose: Artificial intelligence (AI) chatbots, such as ChatGPT and GPT-4, developed by OpenAI, have the potential to revolutionize education. This study explores the potential benefits and challenges of using ChatGPT in education (or "educative AI"). Design/Approach/Methods: This paper proposes a theoretical framework called "IDEE" for educative AI such as using ChatGPT and other generative AI in education, which includes identifying the desired outcomes, determining the appropriate level of automation, ensuring ethical considerations, and evaluating effectiveness. Findings: The benefits of using ChatGPT in education or more generally, educative AI, include a more personalized and efficient learning experience for students as well as easier and faster feedback for teachers. However, challenges such as the untested effectiveness of the technology, limitations in the quality of data, and ethical and safety concerns must also be considered. Originality/Value: This study explored the opportunities and challenges of using ChatGPT in education within the proposed theoretical framework.

This research is part of the Artificial Intelligence Virtual Trainer (AI-VT) project which aims to create a system that can identify the user's skills from a text by means of machine learning. AI-VT is a case-based reasoning learning support system can generate customized exercise lists that are specially adapted to user needs. To attain this outcome, the relevance of the first proposed exercise must be optimized to assist the system in creating personalized user profiles. To solve this problem, this project was designed to include a preliminary testing phase. As a generic tool, AI-VT was designed to be adapted to any field of learning. The most recent application of AI-VT was in the field of computer science specifically in the context of the fundamentals of algorithmic learning. AI-VT can and will also be useful in other disciplines. Developed in Python with the Keras API and the Tensorflow framework, this artificial intelligence-based tool encompasses a supervised learning environment, multi-label text classification techniques and deep neural networks. This paper presents and compares the performance levels of the different models tested on two different data sets in the context of computer programming and algorithms.

Social media have established a new way of communicating and understanding social relationships. At the same time, there are downsides, especially, their use of algorithms that have been built and developed under their umbrella and their potential to alter public opinion. This paper tries to analyse the YouTube recommendation system from the perspectives of reverse engineering and semantic mining. The first result is that, contrary to expectations, the issues do not tend to be extreme from the point of view of polarisation in all cases. Next, and through the study of the selected themes, the results do not offer a clear answer to the proposed hypotheses, since, as has been shown in similar works, the factors that shape the recommendation system are very diverse. In fact, results show that polarising content does not behave in the same way for all the topics analysed, which may indicate the existence of moderators --or corporate actions-- that alter the relationship between the variables. Another contribution is the confirmation that we are dealing with non-linear, but potentially systematic, processes. Nevertheless, the present work opens the door to further academic research on the topic to clarify the unknowns about the role of these algorithms in our societies.

This paper leverages cluster analysis to provide insight into how traditionally underrepresented learners engage with entry-level massive open online courses (MOOCs) intended to lower the barrier to university enrolment, produced by a major research university in the United States. From an initial sample of 260,239 learners, we cluster analyze a subset of data from 29,083 participants who submitted an assignment in one of nine entry-level MOOC courses. Manhattan distance and Gower distance measures are computed based on engagement, achievement, and demographic data. To our knowledge, this marks one of the first such uses of Gower distance to cluster mixed variable data to explore fairness and equity in the MOOC literature. The clusters are derived from CLARA and PAM algorithms, enriched by demographic data, with a particular focus on education level, as well as approximated socioeconomic status (SES) for a smaller subset of learners. Results indicate that learners without a college degree are more likely to be high-performing compared to college-educated learners. Learners from lower SES backgrounds are just as likely to be successful as learners from middle and higher SES backgrounds. While MOOCs have struggled to improve access to learning, more fair and equitable outcomes for traditionally underrepresented learners are possible.

Embedding a pairwise key distribution approach in IoT systems is challenging as IoT devices have limited resources, such as memory, processing power, and battery life. This paper presents a secure and lightweight approach that is applied to IoT devices that are divided into Voronoi clusters. This proposed algorithm comprises XOR and concatenation operations for interactive authentication between the server and the IoT devices. Predominantly, the authentication is carried out by the server. It is observed that the algorithm is resilient against man-in-the-middle attacks, forward secrecy, Denial of Service (DoS) attacks, and offers mutual authentication. It is also observed that the given scheme has low communication and computing overheads compared to some existing methods.

The focus of ethics in learning analytics (LA) frameworks and guidelines is predominantly on procedural elements of data management and accountability. Another, less represented focus is on the duty to act and LA as a moral practice. Data feminism as a critical theoretical approach to data science practices may offer LA research and practitioners a valuable lens through which to consider LA as a moral practice. This paper examines what data feminism can offer the LA community. It identifies critical questions for further developing and enabling a responsible stance in LA research and practice taking one particular case -- algorithmic decision-making -- as a point of departure.

A critical part of supporting the development of students' algorithmic thinking is understanding the challenges that emerge when students engage with algorithmatizing tasks--tasks that require the creation of an algorithm. Knowledge of these challenges can serve as a basis upon which educators can build effective strategies for enhancing students' algorithmic thinking skills. This paper presents three illustrative cases of emergent challenges evident as students grapple with the process of creating an algorithm. The first challenge highlights discrepancies between the method with which students solve a problem and the algorithm they create, and claim would, when implemented, solve the same problem. The second challenge pertains to the persistence of students' normatively incorrect algorithms, despite going through multiple iterations of testing and revising. Finally, the third challenge concerns issues around the use of test problems for supporting students in their creation of generalized algorithms. These three challenges are discussed using student data (illustrative cases) from three different mathematical algorithmatizing tasks. Suggestions are put forth for addressing some of these challenges, with a particular emphasis on practical pedagogical suggestions for cultivating students' mathematical thinking in the context of algorithmatizing tasks.

As we move further into the digital age, the acquisition of digital literacy (DL) and computational thinking (CT) skills is emerging internationally as an essential goal for students in contemporary school curricula. As the world becomes more uncertain and volatile due to impacts of artificial intelligence (AI), international unrest, climate change, global economic instability and unforeseen catastrophes such as the Coronavirus (COVID-19) pandemic, it is necessary to review, revise and refine school education curricula and policies. The issue of what is essential for students to learn, and how they learn it, is of growing importance to international organisations such as the Organisation for Economic Co-operation and Development (OECD) and the United Nations Educational, Scientific and Cultural Organisation (UNESCO) and is emerging as a significant driver for educational reform across the globe. The growing prominence of CT and DL skills across many industry sectors has prompted recent changes to international assessment frameworks such as the Programme for International Student Assessment (PISA) and the Trends in International Mathematics and Science Study (TIMSS). This paper will briefly discuss specific examples of alternative approaches to addressing CT in national curricula for the compulsory years of schooling and explain how CT has been adopted as a driver for mathematics curriculum change in Australia.

This study investigated the influence of CThink4CS2 Module on computational thinking (CT) skills of form four chemistry students. The CThink4CS[superscript]2 Module integrated CT with the Engineering Design Process (EDP) in chemistry class. This study utilized quantitative research methods and quasi-experimental design. Quantitative data were collected using the Computational Thinking Skill Test (CTST) which consisted of algorithmic reasoning, abstraction, decomposition, and pattern recognition constructs. A total of 73 students were in the treatment group (n=39) and control group (n=34). Experimental data were described by means of descriptive analysis and inferential analysis employing two-way MANOVA analysis. The results of the analysis indicated significant differences in CT skills between groups; students in the treatment group demonstrated better results compared to those in the control group. The paper provides insight into the integration of CT and EDP as effective pedagogical strategies for inculcating CT skills.

Formative assessment is about providing and using feedback and diagnostic information. On this basis, further learning or further teaching should be adaptive and, in the best case, optimized. However, this aspect is difficult to implement in reality, as teachers work with a large number of students and the whole process of formative assessment, especially the evaluation of student performance takes a lot of time. To address this problem, this paper presents an approach in which student performance is collected through a concept map and quickly evaluated using Machine Learning techniques. For this purpose, a concept map on the topic of mechanics was developed and used in 14 physics classes in Germany. After the student maps were analysed by two human raters on the basis of a four-level feedback scheme, a supervised Machine Learning algorithm was trained on the data. The results show a very good agreement between the human and Machine Learning evaluation. Based on these results, an embedding in everyday school life is conceivable, especially as support for teachers. In this way, the teacher can use and interpret the automatic evaluation and use it in the classroom.

Computational thinking (CT) is an important 21st-century skill. This paper aims at investigating predictors of CT self-efficacy among high-school students. The hypothesized predictors are grouped into three areas: (1) student characteristics, (2) home environment, and (3) learning opportunities. CT self-efficacy is measured with the Computational Thinking Scales (CTS) that comprises five dimensions: creativity, algorithmic thinking, cooperativity, critical thinking, and problem solving. N = 202 high-school students act as the sample, linear regression as the analysis method. The best prediction is possible for algorithmic thinking (R[superscript 2] = 0.511). For cooperativity, the explanatory power of our model it is weak (R[superscript 2] = 0.146). Across all five CTS dimensions, CT self-concept is the best predictor for CT self-efficacy. [For the full proceedings, see ED636095.]

Students organize and manage their own learning time, choosing when, what, and how to study due to the flexibility of online learning. Each person has unique learning habits that define their behaviours and distinguish them from others. To investigate the temporal behaviour of students in online learning environments, we seek to identify suitable time-windows that could be used to investigate their temporal behaviour. First, we present a novel perspective for identifying different types of sessions based on individual needs. The majority of previous works address this issue by establishing an arbitrary session timeout threshold. In this paper, we propose an algorithm for determining the optimal threshold for a given session. Second, we use data-driven methods to support investigators in determining time-windows based on the identified sessions. To this end, we developed a visual tool that assists data scientists and researchers to determine the optimal settings for session identification and locating suitable time-windows.

Cognitive Diagnosis Models (CDMs) are a special family of discrete latent variable models that are widely used in educational and psychological measurement. A key component of CDMs is the Q-matrix characterizing the dependence structure between the items and the latent attributes. Additionally, researchers also assume in many applications certain hierarchical structures among the latent attributes to characterize their dependence. In most CDM applications, the attribute--attribute hierarchical structures, the item-attribute Q-matrix, the item-level diagnostic models, as well as the number of latent attributes, need to be fully or partially pre-specified, which however may be subjective and misspecified as noted by many recent studies. This paper considers the problem of jointly learning these latent and hierarchical structures in CDMs from observed data with minimal model assumptions. Specifically, a penalized likelihood approach is proposed to select the number of attributes and estimate the latent and hierarchical structures simultaneously. An expectation-maximization (EM) algorithm is developed for efficient computation, and statistical consistency theory is also established under mild conditions. The good performance of the proposed method is illustrated by simulation studies and real data applications in educational assessment.

Machine learning (ML), a form of artificial intelligence (AI), is being increasingly employed in neurology. Reported performance metrics often match or exceed the efficiency of average clinicians. The neurologist is easily baffled by the underlying concepts and terminologies associated with ML studies. The superlative performance metrics of ML algorithms often hide the opaque nature of its inner workings. Questions regarding ML model's interpretability and reproducibility of its results in real-world scenarios, need emphasis. Given an abundance of time and information, the expert clinician should be able to deliver comparable predictions to ML models, a useful benchmark while evaluating its performance. Predictive performance metrics of ML models should not be confused with causal inference between its input and output. ML and clinical gestalt should compete in a randomized controlled trial before they can complement each other for screening, triaging, providing second opinions and modifying treatment. [ABSTRACT FROM AUTHOR]