Search

Adaptive online learning can be realized through the evaluation of the learning process. Monitoring and supervising learners' cognitive levels and adjusting learning strategies can increasingly improve the quality of online learning. This analysis is made possible by real-time measurement of learners' cognitive levels during the online learning process. However, most of the currently used techniques for evaluating cognitive levels rely on labour-intensive and time-consuming manual coding. In this study, we explore the machine learning (ML) algorithms and taxonomy of Bloom's cognitive levels to explore features that affect learner's cognitive level in online assessments and the ability to automatically predict learner's cognitive level and thus, come up with a recommendation or pedagogical intervention to improve learner's acquisition. The analysis of 15,182 learners' assessments of a specific learning concept affirms the effectiveness of our approach. We attain an accuracy of 82.21% using ML algorithms. These results are very encouraging and have implications for how automated cognitive-level analysis tools for online learning will be developed in the future.

This study was carried out to develop a test to assess algorithmic thinking skills. To this end, the twelve steps suggested by Downing (2006) were adopted. Throughout the test development, 24 middle school sixth-grade students and eight experts in different areas took part as needed in the tasks on the project. The test was given to 252 students attending the sixth grade who were selected through purposeful sampling. The content validity of the test was ensured by means of obtaining expert opinion, whereas the construct validity was ensured by performing an independent sample t-test on the difference between the lower and upper groups. As a result, the algorithmic thinking skills assessment test was finalized with 22 main items and 2 sample items, totalling 24 items. The KR-20 reliability analysis proved a quite reliable test based on the reliability coefficient of 0.83. As mentioned earlier, the independent sample t-test was applied to the difference of lower and upper groups for construct validation of the test. It was seen that the test items are significant in discriminating the students in the lower and upper groups (p<0.01).

The study focuses on the identification of relationships and/or rules between computational thinking (CT) concepts among the undergraduate students of Applied Informatics due to their attitudes towards mathematics. We analyze three CT concepts -- decomposition, pattern recognition, and algorithmic thinking. We assume that students who have a closer relationship with mathematics, a positive attitude towards mathematics, have better developed CT. We conducted the experiment during the 2022/2023 academic year on the Information Coding and Displaying subject. The results indicate that those students who have no relationship to mathematics, negative attitude towards mathematics, have no problem with decomposition and pattern recognition, but without significant algorithmic thinking. On the other hand, students who have a close relationship with mathematics are also able to decompose or recognize patterns, but moreover they have shown algorithmic thinking. The contribution of the study comprises the identification of relationships and/or patterns of computational thinking concepts among students who have a relationship to mathematics, who cannot assess their relationship to mathematics, as well as among students who have no relationship to mathematics. Our results indicate a different occurrence of computational thinking concepts as well as links and/or relationships between them.

This study applies the Apriori algorithm to analyse patterns, syntactic structures, and thematic clusters in Chinese studies data from various genres. This study aims to identify recurring linguistic elements in order to shed light on the dynamic nature of the Chinese language across different contexts and time periods. The Apriori algorithm is used to identify frequent item sets and establish associations between linguistic constructs in large datasets (over 20 years). This study examines the complexity of the Chinese language by analysing co-occurrence patterns, syntactic tendencies, and thematic categorizations. This study examines the evolution of language, regional word choices, and cultural nuances. Thematic clusters and sensory associations establish the relationship between language and culture. The study of Chinese language patterns and cultural implications utilises data to advance computational linguistics and theory. Computational models prioritize cultural and historical context analysis for more comprehensive language processing. Theoretical implications help researchers understand language evolution and culture, while practical implications improve language technology tools. The conclusion provides support for research in computational linguistics, cultural studies, and linguistic theory-based holistic language analysis and application.

Artificial Intelligence (AI) has emerged as a prominent area of investigation in the field of education. Also, perceptions, challenges and threats of AI among university professors show notable variations. This study explores university professors' perspectives regarding AI, including their familiarity with AI, its educational impacts, challenges associated with its implementation, and perceived threats. To achieve this, a survey was administered to 65 university professors from diverse Egyptian institutions, both state and private. Subsequent statistical analyses were conducted to treat the collected data. The outcomes of these analyses reveal that university professors possess varying degrees of familiarity with AI. Despite this, they view AI as a valuable educational tool. The study identifies several challenges hindering AI adoption, encompassing difficulties in comprehending and interpreting AI algorithmic outcomes, the intricate autonomy of AI systems, financial implications of implementation, and concerns regarding data privacy. Additionally, the study identifies apprehensions pertaining to AI''s influence on professors' skills, potential dehumanization of pedagogy, adverse effects on students, and the potential obsolescence of professors. These findings bear implications for the integration of AI in educational contexts, highlighting the necessity for further exploration in this evolving field.

This study explores the transformative potential of Generative AI (GenAI) and ChatBots in educational interaction, communication, and the broader implications of human-GenAI collaboration. By examining the related literature through data mining and analytical methods, the paper identifies three main research themes: the revolutionary role of GenAI-powered ChatBots in educational interactions, their capability to enrich social learning, and their dual role as both support and assistance within educational settings. This research further highlights the impact of human-GenAI interaction in education from social, psychological, and cultural perspectives, focusing on social presence as a fundamental component of the teaching and learning process. It discusses the integration of GenAI and ChatBots into education and considers whether this marks the dawn of an algorithmic renaissance that elevates educational experiences or an apocalypse that threatens the very essence of human learning and interaction.

In learning mathematics in the 21st century and after the COVID-19 pandemic, a multimodal role in the process of drawing conclusions involving mathematical symbols or signs is very much needed. This process is called multimodal semiotic reasoning. This research aims to study the literature on multimodal semiotic reasoning research articles. Until April 2023, there were 32 articles on the keyword multimodal semiotic reasoning with the subject of mathematics in the Scopus, Science Direct, and ERIC databases. Furthermore, it is identified based on a) suitability of criteria; b) the similarity of the articles in the three databases. Finally obtained 24 articles that meet the requirements for further research. A literature review on multimodal semiotic reasoning is explored based on a) kinds of mathematical reasoning involving symbols and signs using multimodal, b) multimodal used in research, c) the most researched mathematical concepts and the mathematization process that occurs, d) the role of multimodal semiotic reasoning in mathematics education. The results showed that a) the most dominant type of mathematical reasoning involving objects, symbols, and signs using multimodal was Algorithmic Reasoning (AR), b) the most multimodal source was verbal answers, c) the concept of geometry through a process of vertical-horizontal mathematization was the most researched, and 4) the biggest role of multimodal semiotic reasoning for the benefit of knowledge of mathematics education. In conclusion, multimodal semiotic reasoning has had a major impact on mathematics education and research on this multimodal semiotic reasoning will continue to thrive in the future. [Note: The publication year (2023) shown in the citation on the PDF is incorrect. The correct publication year is 2024.]

This manuscript reports the results of an investigation of the simultaneous prompting procedure to teach a five-step algorithm for solving three addition basic facts. Four preschool students receiving Tier 1 services in their school's multi-tier system of supports (MTSS) framework participated. A multiple probe across participants single case design was employed, and a visual analysis of the data indicate a functional relationship between the independent and dependent variable. Three of the participants achieved the mastery criterion and maintained the targeted learning outcome for 2-6 weeks. These participants applied the five-step algorithm to three sets of novel addition facts and acquired incidental information consisting of mathematics terminology. The fourth participant voluntarily discontinued her involvement in the study but, prior to her withdrawal, her data were similar to that of the other three participants. Social validity data from the participants' teachers confirmed their satisfaction with the study's procedures and participants' learning outcomes. The results extend those from related research and validate simultaneous prompting's effectiveness in teaching mathematics to students receiving Tier 1 and Tier 3 services in an MTSS framework. Topics for future research are discussed, and include investigating the effectiveness of simultaneous prompting to teach mathematics across students receiving services through all MTSS tiers.

The emergence of artificial intelligence (AI) in academia has prompted various debates on the uses, threats, and limitations of tools that can create text for numerous academic purposes. Critics argue that these advancements may provide opportunities for cheating and plagiarism and even replace the art of writing entirely. To reclaim the creativity and depth that academic writing holds, we propose both an innovative approach to safeguard the creativity and depth of academic writing and a scaffolded way to enhance success in terms of authenticity for the author and, by extension, meaning for the reader. This novel conceptual algorithmic trickle filter model aims to inform successful academic writing and embody the writer's agency--a task too sophisticated for current AI tools. Our model provides a scaffolded decision-making process in a highly personal, flexible, and iterative individual writing development tool applied in a health-conscious way. We position this model as a step towards a pedagogic paradigm shift in reclaiming academic writing that, rather than competing with AI, doubles down on the personal self-evaluative aspects that academic writing offers both author and reader.

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.

The rapid development of education informatization has accumulated a large amount of data for learning analytics, and adopting educational data mining to find new patterns of data, develop new algorithms and models, and apply known predictive models to the teaching system to improve learning is the challenge and vision of the education field in the era of big data. Learning intervention, as a core concept of learning analytics, refers to the purposeful and planned adoption of direct or indirect strategies or behaviors based on tracking learning behaviors and integrating information about learners' characteristics to give learners personalized guidance and assistance in order to help learners break through the status quo of learning difficulties and improve their learning abilities, so as to achieve tailored teaching. In this study, data mining was conducted on the performance records of students on math problems in an online learning system, and a learning intervention suggestion prediction model was constructed on the basis of decision tree algorithm using Python, with a view to understanding the effectiveness, willingness, style, and other characteristics of the learners' online learning through the analysis results, providing personalized guidance to students, and enabling teachers to intervene with at-risk students and successfully complete the teaching goals. It was found that the most significant impact of the learning intervention advice provided to learners was the number of hints they sought during the learning process, and that learners who needed to be "intervened" or "monitored" could be categorized into two groups: independent inefficient and dependent inefficient according to the model. Therefore, teachers or adaptive learning systems should intervene in a timely and appropriate way for different types of learning crisis groups to solve the problems of poor learning performance, insufficient commitment to learning, poor learning habits, low participation in learning, low self-efficacy and other problems of learners in different learning scenarios.

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.

We propose a novel approach to address the issue of college student attrition by developing a hybrid model that combines a structural neural network with a piecewise exponential model. This hybrid model not only shows the potential to robustly identify students who are at high risk of dropout, but also provides insights into which factors are most influential in dropout prediction. To evaluate its effectiveness, we compared the predictive performance of our hybrid model with two other survival analysis models: the piecewise exponential model and a hybrid model combining a fully-connected neural network with a piecewise exponential model. Additionally, we compared it to five other cross-sectional models: Ridge Logistic Regression, Lasso Logistic Regression, CART decision tree, Random Forest, and XGBoost decision tree. Our findings demonstrate that the hybrid model outperforms or performs comparably to the other models when predicting dropout among students at the University of Delaware in Spring 2020, Spring 2021, and Spring 2022. Moreover, by categorizing predictors into three distinct groups--academic, economic, and social-demographic--we discovered that academic predictors play a prominent role in distinguishing between dropout and retained students, while other predictors may indirectly influence predictions by impacting academic variables. Consequently, we recommend implementing an intervention program aimed at identifying at-risk students based on their academic performance and activities, with additional consideration for economic and social-demographic factors in customized intervention plans.

In higher education, academic advising is crucial to students' decision-making. Data-driven models can benefit students in making informed decisions by providing insightful recommendations for completing their degrees. To suggest courses for the upcoming semester, various course recommendation models have been proposed in the literature using different data mining techniques and machine learning algorithms utilizing different data types. One important aspect of the data is that usually, courses taken together in a semester fit well with each other. If there is no correlation between the co-taken courses, students may find it more difficult to handle the workload. Based on this insight, we propose using session-based approaches to recommend a set of well-suited courses for the upcoming semester. We test three session-based course recommendation models, two based on neural networks (CourseBEACON and CourseDREAM) and one on tensor factorization (TF-CoC). Additionally, we propose a post-processing approach to adjust the recommendation scores of any base course recommender to promote related courses. Using metrics capturing different aspects of the recommendation quality, our experimental evaluation shows that session-based methods outperform existing popularity-based, association-based, similarity-based, factorization-based, neural networks-based, and Markov chain-based recommendation approaches. Effective course recommendations can result in improved student advising, which, in turn, can improve student performance, decrease dropout rates, and a more positive overall student experience and satisfaction.

Predicting and improving the academic achievement of university students is a multifactorial problem. Considering the low success rates and high dropout rates, particularly in open education programs characterized by mass enrollment, academic success is an important research area with its causes and consequences. This study aimed to solve a classification problem (successful or unsuccessful), predict students' academic success, and identify those at risk. The primary objective was to predict the academic success status with 26,708 students enrolled in Istanbul University open and distance education programs between 2011 and 2017. Predictions were based demographic data and success grades in Turkish, Atatürk's Principles and History of Revolution, English, and Disaster Culture courses. The study utilized classification models from supervised learning algorithms and was conducted using the SPSS Modeler 18 program. Initially, the data was divided into 70% training and 30% test data. Then, models were constructed by using Random Forest, Tree-AS, C&RT, C5.0, CHAID, QUEST, Naive Bayes, Logistic Regression, NeuralNet, and SVM algorithms. Model performances were compared according to accuracy, sensitivity, specificity, F1 score, positive predictive value, negative predictive value, and Matthews Correlation Coefficient criteria. The C&RT model demonstrated the best performance, achieving the highest specificity value of 0.915.

Hollywood screenwriters worry about Artificial Intelligence (AI) replacements taking over their jobs. Famous museums litigate to protect their art from AI infringement. A major retailer scraps a machine-learning based recruitment program that was biased against women. These are just a few examples of how AI is affecting the world of work, learning, and living. MIS and computer science students are among the professional groups who are embarking into careers with nebulous frontiers obscured by the outcroppings brought on by AI. Computer Science and Information System curriculum task forces have recognized the increasing ethical and professional implications developers' work can have beyond the scope of the programmers' code. In this article, the authors examine the professional, legal, and ethical implications of copyrights and algorithmic bias resulting from development of AI-enhanced applications and offer suggestions for addressing these topics in courses considering changes to the CC2020 and IS2020 Model Curriculum frameworks.

The cyber security education market has grownup exponentially, with a CAGR of 13.9 % as reported by Data Intelo. The report published by the World Economic Forum 2023 indicates a shortfall of 2.27 million cyber security experts in 2021 across different roles and hence manifest that Skill-based cyber security education is the need of the hour. Cybersecurity as a field has evolved as a multi-discipline, multi-stakeholder and multi-role discipline. Therefore, the need to address formal education with an outcome-based philosophy is imperative to address for a wider audience with varied past training in their formal education. With the Internet becoming an essential part of human life, providing security of data passed over the Internet is becoming increasingly crucial. Therefore, the role in the organization which is quite demanding is to have expertise in handling and configuring network security, a subdomain of cyber security as a priority area. The rapid increase in the network attack landscape is constantly demanding monitoring of network attacks as well as a need to promote collaborative R&D and education in the sphere of cyber security due to a shortage of skilled resources. Platform-based education is a potential direction to achieve the objective and address the skill gap required in cyber security. The present research proposes a comprehensive web-based platform that can be used to communicate, collaborate and practice various use cases in the domain of network intrusion detection tools using machine learning algorithms and to evaluate user experience. The proposed platform CySecLearn is a collection of various functionalities and features that ensure experiential learning, will help the learner to develop critical thinking and expertise in network security and promote digital literacy in the domain of cyber security.

Previous studies have found positive effects of Game-Based Learning for mathematics. While most studies assume that this effect is explained by the presence of flow/immersion during games, this has not yet been established. The aim of the current study is to verify if immersion indeed is associated with mathematical skills improvement when using a Game-Based Learning intervention. This was tested among 59 Greek high school students, using authentic design. After having received a traditional education module, the students were tested and then engaged for four weeks in a desktop-based 3D Virtual Learning Environment where they could play mathematic minigames. They were subsequently re-tested to verify if they showed a significant increase in mathematical skills. The students showed an improvement in their mathematical skills (Cohen's d = 1.26), with significant results for functions, geometry, and thinking skills and methods. On the individual level, about half of the students showed a 10% increase in one of the domains (numbers & calculations, functions, geometry, thinking skills and methods, and algorithms and number theory). Immersion was found to be reflected by engagement and presence, but neither one of these aspects was associated with mathematical achievement after the intervention. It is concluded that Game-Based Learning is an effective approach to increasing mathematical skills, yet the underlying mechanisms are not yet understood. The authors discuss several alternative mechanisms based on the literature that can be verified in future studies.

The use of a pedagogical approach mediated transfer with the bridging method has been successful in facilitating the transitions from block-based to text-based programming languages. Nevertheless, there is a lack of research addressing the impact of this transfer on programming misconceptions during the transition. The way programming concepts are taught to K-12 learners can later result in misconceptions for adult learners. The main objective was to examine the impact of mediated transfer using the bridging method pedagogical approach on the prevalence of programming misconceptions. We conducted a quasi-experimental study in school settings during informatics (computer science) classes among 163 sixth-grade students. The control group received traditional programming lectures using the text-based programming language, Python. Conversely, the experimental group utilized a mediated transfer pedagogical approach by starting with the block-based programming language MakeCode for micro:bit before transitioning to the text-based Python. Our findings indicate that the experimental group significantly reduced programming misconceptions in fundamental programming concepts: variables, sequencing, selection, and loops - compared to the control group. This suggests that the use of block-based programming language as an initial step in programming education, followed by a structured transition to text-based programming language, can effectively mitigate common misconceptions among K-12 learners.

This paper develops a spreadsheet simulation methodology for teaching simulation and performance analysis of priority queues with multiple servers, without resorting to macros, add-ins, or array formula. The approach is made possible by a "single overtaking" simplifying assumption under which any lower-priority customer may be passed in line by at most one higher-priority customer. By increasing the number of overtaking customers, one at a time, the simulation model is extended to the "multiovertaking" case. These simplifying assumptions make such spreadsheet simulations (of more complex queuing networks) accessible to students, and so the paper includes teaching and learning strategies for the classroom. Performance analysis of single-overtaking versus multiovertaking polices is included.

Addressing cognitive disparities has become a paramount concern in computational thinking (CT) education. The intricate and nuanced relationships between CT and cognitive variations emphasize the needs to accommodate diverse cognitive profiles when fostering CT skills, recognizing that these cognitive functions can manifest as either strengths or limitations in different students. Consequently, understanding the connections between students' cognitive functions and CT skills assumes pivotal importance in the design of personalized instructional strategies for CT. Despite a general consideration of learning variability in CT education, empirical insights exploring the correlation between cognitive skills and CT competencies remain notably scarce. This study endeavors to bridge this research gap by investigating the links between executive functions and CT skills, as well as the associations between their sub-dimensions. The results reveal a statistically significant correlation coefficient of 0.452 between these two domains, underscoring the notable connection between executive functions and CT abilities. Furthermore, the sub-dimensional analysis offers a comprehensive understanding of how specific executive functions uniquely contribute to certain CT skills. In light of these findings, this research offers a promising pathway for the development of tailored CT education programs that can cater to the unique needs of each individual, ultimately facilitating inclusive CT programs and making significant contributions to broaden STEM education and future workforce.

The impact of text length on the estimation of lexical diversity has captured the attention of the scientific community for more than a century. Numerous indices have been proposed, and many studies have been conducted to evaluate them, but the problem remains. This methodological review provides a critical analysis not only of the most commonly used indices in language learning studies, but also of the length problem itself, as well as of the methodology for evaluating the proposed solutions. Analysis of three data sets of texts produced by English language learners revealed that indices that reduce all texts to the same length using a probabilistic or an algorithmic approach solve the length-dependency problem; however, all these indices failed to address the second problem, which is their sensitivity to the parameter that determines the length to which the texts are reduced. The paper concludes with recommendations for optimizing lexical diversity analysis.

Large language models (LLMs) are increasingly adopted in educational contexts to provide personalized support to students and teachers. The unprecedented capacity of LLM-based applications to understand and generate natural language can potentially improve instructional effectiveness and learning outcomes, but the integration of LLMs in education technology has renewed concerns over algorithmic bias, which may exacerbate educational inequalities. Building on prior work that mapped the traditional machine learning life cycle, we provide a framework of the LLM life cycle from the initial development of LLMs to customizing pre-trained models for various applications in educational settings. We explain each step in the LLM life cycle and identify potential sources of bias that may arise in the context of education. We discuss why current measures of bias from traditional machine learning fail to transfer to LLM-generated text (e.g., tutoring conversations) because text encodings are high-dimensional, there can be multiple correct responses, and tailoring responses may be pedagogically desirable rather than unfair. The proposed framework clarifies the complex nature of bias in LLM applications and provides practical guidance for their evaluation to promote educational equity.

In e-learning environments, most adaptive systems do not consider the learner's emotional state when recommending activities for learning difficulties, blockages, or demotivation. In this paper, we propose a new approach of emotion-based adaptation in e-learning environments. The system will allow recommendation resources/activities to motivate and support the learner in learning. Our first contribution is modeling the learner's emotion by exploiting the facial expressions generated during the pedagogical activities. For this purpose, a probability-based emotion quantification algorithm has been proposed. To recommend support resources, we presented our adaptation approach that leverages a set of adopted adaptation criteria, where the weighting of these criteria differs from one support resource to another. Five experiments aimed at validating the approach were conducted on two groups of students (test and control groups). The results show our approach's impact on improving the learner's cognitive level, engagement time, and motivation.

In the context of our rapidly digitizing society, computational thinking stands out as an essential attribute for cultivating aptitude and expertise. Through the prism of computational thinking, learners are more adeptly positioned to dissect and navigate real-world challenges, poising them effectively to meet the exigencies of future societal landscapes. The paradigm of project-based learning, heralded as a potent educational methodology, significantly amplifies students' prowess in computational thinking. To delineate the nuanced interplay between project-based learning and its augmentation of computational thinking, we engaged in meta-analytical investigations, synthesizing outcomes from 31 distinct external experimental and quasi-experimental studies. Our analytical journey underscored that project-based learning markedly elevates students' competencies across five critical spectra: innovation, collaboration, critical analysis, algorithmic cognition, and problem resolution. Dissecting its impact across the K12 continuum reveals that primary students predominantly hone their collaborative acumen; middle-grade learners witness a pronounced ascent in creativity, critical, and algorithmic faculties, while the apex of high school education primarily sharpens critical analytical prowess. Thus, educators sculpting the contours of project-based learning should be calibrate their strategies, ensuring they resonate with the developmental and age-specific nuances of their audience, to fully unleash its latent potential in nurturing computational thinking.

In the current educational landscape, where large amounts of data are being produced by institutions, Educational Data Mining (EDM) emerges as a critical discipline that plays a crucial role in extracting knowledge from this data to help academic policymakers make decisions. EDM has a primary focus on predicting students' academic performance. Numerous studies have been conducted for this purpose, but they are plagued by challenges including limited dataset size, disparities in grade distributions, and feature selection issues. This paper introduces a Machine Learning (ML) based method for the early prediction of bachelor students' final academic grade as well as drop-out cases. It focuses on identifying, from the first semester of study, the students requiring specific attention because of their academic weaknesses. The research employs nine classification models on students' data from a Saudi university, subsequently implementing a majority voting algorithm. The experimental outcomes are noteworthy, with the Extra Trees (ET) algorithm achieving a promising accuracy of 82.8% and the Majority Voting (MV) model outperforming all existing models by an accuracy reaching 92.7%. Moreover, the study identifies the factors exerting the greatest impact on students' academic performance, which belong to the three considered feature types: demographic, pre-admission, and academic.

Grouped and right-censored (GRC) counts have been used in a wide range of attitudinal and behavioural surveys yet they cannot be readily analyzed or assessed by conventional statistical models. This study develops a unified regression framework for the design and optimality of GRC counts in surveys. To process infinitely many grouping schemes for the optimum design, we propose a new two-stage algorithm, the Fisher Information Maximizer (FIM), which utilizes estimates from generalized linear models to find a global optimal grouping scheme among all possible N-group schemes. After we define, decompose, and calculate different types of regressor-specific design errors, our analyses from both simulation and empirical examples suggest that: 1) the optimum design of GRC counts is able to reduce the grouping error to zero, 2) the performance of modified Poisson estimators using GRC counts can be comparable to that of Poisson regression, and 3) the optimum design is usually able to achieve the same estimation efficiency with a smaller sample size.

The estimation of RxC ecological inference contingency tables from aggregate data is one of the most salient and challenging problems in the field of quantitative social sciences, with major solutions proposed from both the ecological regression and the mathematical programming frameworks. In recent decades, there has been a drive to find solutions stemming from the former, with the latter being less active. From the mathematical programming framework, this paper suggests a new direction for tackling this problem. For the first time in the literature, a procedure based on linear programming is proposed to attain estimates of local contingency tables. Based on this and the homogeneity hypothesis, we suggest two new ecological inference algorithms. These two new algorithms represent an important step forward in the ecological inference mathematical programming literature. In addition to generating estimates for local ecological inference contingency tables and amending the tendency to produce extreme transfer probability estimates previously observed in other mathematical programming procedures, these two new algorithms prove to be quite competitive and more accurate than the current linear programming baseline algorithm. Their accuracy is assessed using a unique dataset with almost 500 elections, where the real transfer matrices are known, and their sensitivity to assumptions and limitations are gauged through an extensive simulation study. The new algorithms place the linear programming approach once again in a prominent position in the ecological inference toolkit. Interested readers can use these new algorithms easily with the aid of the R package lphom.

The use of algorithms and automation of public services is not new, but in recent years there has been a step change in processing power and a decrease in the price of these technologies, which means we are seeing more widespread use. These advances are reframing our perception of what matters in ways that impact the ethical dimensions of day-to-day life. In turn, these changes challenge long-standing assumptions about public service ethics and how it is taught. In this multidisciplinary authored paper, we argue that public service leaders must be attentive to ethical questions that converge around adopting "data-driven" techniques, including algorithmic decision-making. Algorithmic and technology focused ethics question assumptions about the current deficits within public service ethics pedagogy in public service programs and university programs and the future direction of the discipline. To do so raises longstanding but neglected questions about the public services' role in the state and recovering what Rohr refers to as the 'ethics of the office.' This, we argue, will have implications for teaching public service ethics.

In recent years, computing education researchers have investigated the impact of problem context on students' learning and programming performance. This work continues the investigation motivated, in part, by cognitive load theory and educational research in computer science and other disciplines. The results of this study could help inform computing assessment design. If the context and authenticity of a programming problem aid student performance then, instructors' time in creating appropriately contextualized programming problem descriptions is time well-spent. On the other hand, if the context of a programming problem hinders performance, then instructors should leave it out of programming problems. Recent studies investigating the impact of programming problem context on student success have arrived at different conclusions. Presented here is a series of experiments, conducted over 3 years, investigating the impact of context on novice programmers' success in algorithmic programming assignments using three contextualized tasks and their generic counterparts. This experiment series also looked into the possibility of "authenticity" as a factor affecting performance. Common sense would suggest that a student would perform better on a problem if they understood or cared about it. Contextualization could provide authenticity and authenticity could provide interest. Research suggests that perceiving a problem as authentic has a positive effect on engagement and learning. Alternatively, if a problem is "just an abstract set of numbers", it may be harder to make sense of the details and the lack of context could consequently contribute additional cognitive load to an already challenging assignment. The results of this study show that assignment context and problem context authenticity have no effect on the performance of novice programmers. We think, however, that contextualization could be worth investing in to support students' interest in computing. Additional implications of the results suggest that instructors can assign equivalent versions of the same problem in varied contexts to suit their students' interests without worrying if the context will hinder performance.

Multidimensional achievement tests are recently gaining more importance in educational and psychological measurements. For example, multidimensional diagnostic tests can help students to determine which particular domain of knowledge they need to improve for better performance. To estimate the characteristics of candidate items (calibration) for future multidimensional achievement tests, we use optimal design theory. We generalize a previously developed exchange algorithm for optimal design computation to the multidimensional setting. We also develop an asymptotic theorem saying which item should be calibrated by examinees with extreme abilities. For several examples, we compute the optimal design numerically with the exchange algorithm. We see clear structures in these results and explain them using the asymptotic theorem. Moreover, we investigate the performance of the optimal design in a simulation study.

Most existing diagnostic models are developed to detect whether students have mastered a set of skills of interest, but few have focused on identifying what scientific misconceptions students possess. This article developed a general dual-purpose model for simultaneously estimating students' overall ability and the presence and absence of misconceptions. The expectation-maximization algorithm was developed to estimate the model parameters. A simulation study was conducted to evaluate to what extent the parameters can be accurately recovered under varied conditions. A set of real data in science education was also analyzed to examine the viability of the proposed model in practice.

Simulation is a powerful approach that plays a significant role in science and technology. Computational models that simulate learner interactions and data hold great promise for educational technology as well. Amongst others, simulated learners can be used for teacher training, for generating and evaluating hypotheses on human learning, for developing adaptive learning algorithms, for building virtual worlds in which students can practice collaboration skills with simulated pals, and for testing learning environments. This paper provides the first systematic literature review on simulated learners in the broad area of artificial intelligence in education and related fields, focusing on the decade 2010-19. We analyze the trends regarding the use of simulated learners in educational technology within this decade, the purposes for which simulated learners are being used, and how the validity of the simulated learners is assessed. We find that simulated learner models tend to represent only narrow aspects of student learning. And, surprisingly, we also find that almost half of the studies using simulated learners do not provide "any" evidence that their modeling addresses the most fundamental question in simulation design -- is the model valid? This poses a threat to the reliability of results that are based on these models. Based on our findings, we propose that future research should focus on developing more complete simulated learner models. To validate these models, we suggest a standard and universal criterion, which is based on the lasting idea of Turing's Test. We discuss the properties of this test and its potential to move the field of simulated learners forward.

Knowledge maps have been widely used in knowledge elicitation and representation to evaluate and guide students' learning. To effectively evaluate maps, instructors must select the most informative map features that capture students' knowledge constructs. However, there is currently no clear and consistent criteria to select such features, as empirical studies continue to reflect the (implicit) preferences of scholars. This is challenging for instructors, who may thus ignore critical aspects of a map and/or waste their efforts by examining highly correlated features. To address the research gap of selecting informative graph metrics to assess knowledge maps, this study adopts the machine learning technique of Unsupervised Feature Selection (UFS). Specifically, we extract 12 features used in the prior literature on map assessment (e.g., density, diameter) and use 8 UFS algorithms to rank their importance. By using 202 maps originating from four case studies, we identify features that are generally (un)informative and observe nuances due to context (e.g., learning task, participant profiles). Results suggest that features commonly reported (e.g., number of edges) may not be as informative as less commonly examined aspects (e.g., average degree). Differences exist between maps: the diameter is valuable when learners produce maps from detailed studies, but less informative when maps are directly elicited from the learners' perspectives. The 8 UFS algorithms show five distinct ways to rank features in maps, hence future works may elicit the preferences of instructors for grading and map these preferences to an algorithmic approach (i.e., UFS) that produces a desired ranking.

Artificial intelligence in higher education is becoming more prevalent as it promises improvements and acceleration of administrative processes concerning student support, aiming for increasing student success and graduation rates. For instance, Academic Performance Prediction (APP) provides individual feedback and serves as the foundation for distributing student support measures. However, the use of APP with all its challenges (e.g., inherent biases) significantly impacts the future prospects of young adults. Therefore, it is important to weigh the opportunities and risks of such systems carefully and involve affected students in the development phase. This study addresses students' fairness perceptions of the distribution of support measures based on an APP system. First, we examine how students evaluate three different distributive justice norms, namely, "equality," "equity," and "need." Second, we investigate whether fairness perceptions differ between APP based on human or algorithmic decision-making, and third, we address whether evaluations differ between students studying science, technology, engineering, and math (STEM) or social sciences, humanities, and the arts for people and the economy (SHAPE), respectively. To this end, we conducted a cross-sectional survey with a 2 X 3 factorial design among n = 1378 German students, in which we utilized the distinct distribution norms and decision-making agents as design factors. Our findings suggest that students prefer an equality-based distribution of support measures, and this preference is not influenced by whether APP is based on human or algorithmic decision-making. Moreover, the field of study does not influence the fairness perception, except that students of STEM subjects evaluate a distribution based on the need norm as more fair than students of SHAPE subjects. Based on these findings, higher education institutions should prioritize student-centric decisions when considering APP, weigh the actual need against potential risks, and establish continuous feedback through ongoing consultation with all stakeholders.

Cognitive diagnosis models (CDMs) are the assessment tools that provide valuable formative feedback about skill mastery at both the individual and population level. Recent work has explored the performance of CDMs with small sample sizes but has focused solely on the estimates of individual profiles. The current research focuses on obtaining accurate estimates of skill mastery at the population level. We introduce a novel algorithm (bagging algorithm for deterministic inputs noisy "and" gate) that is inspired by ensemble learning methods in the machine learning literature and produces more stable and accurate estimates of the population skill mastery profile distribution for small sample sizes. Using both simulated data and real data from the Examination for the Certificate of Proficiency in English, we demonstrate that the proposed method outperforms other methods on several metrics in a wide variety of scenarios.

The cross-classified data structure is ubiquitous in education, psychology, and health outcome sciences. In these areas, assessment instruments that are made up of multiple items are frequently used to measure latent constructs. The presence of both the cross-classified structure and multivariate categorical outcomes leads to the so-called item-level data with cross-classified structure. An example of such data structure is the routinely collected student evaluation of teaching (SET) data. Motivated by the lack of research on multilevel IRT modeling with crossed random effects and the need of an approach that can properly handle SET data, this study proposed a cross-classified IRT model, which takes into account both the cross-classified data structure and properties of multiple items in an assessment instrument. A new variant of the Metropolis-Hastings Robbins-Monro (MH-RM) algorithm was introduced to address the computational complexities in estimating the proposed model. A preliminary simulation study was conducted to evaluate the performance of the algorithm for fitting the proposed model to data. The results indicated that model parameters were well recovered. The proposed model was also applied to SET data collected at a large public university to answer empirical research questions. Limitations and future research directions were discussed.

Identifying items with differential item functioning (DIF) in an assessment is a crucial step for achieving equitable measurement. One critical issue that has not been fully addressed with existing studies is how DIF items can be detected when data are multilevel. In the present study, we introduced a Lord's Wald X[superscript 2] test-based procedure for detecting both uniform and non-uniform DIF with polytomous items in the presence of the ubiquitous multilevel data structure. The proposed approach is a multilevel extension of a two-stage procedure, which identifies anchor items in its first stage and formally evaluates candidate items in the second stage. We applied the Metropolis-Hastings Robbins-Monro (MH-RM) algorithm to estimate multilevel polytomous item response theory (IRT) models and to obtain accurate covariance matrices. To evaluate the performance of the proposed approach, we conducted a preliminary simulation study that considered various conditions to mimic real-world scenarios. The simulation results indicated that the proposed approach has great power for identifying DIF items and well controls the Type I error rate. Limitations and future research directions were also discussed.

Forced-choice (FC) measures have been widely used in many personality or attitude tests as an alternative to rating scales, which employ comparative rather than absolute judgments. Several response biases, such as social desirability, response styles, and acquiescence bias, can be reduced effectively. Another type of data linked with comparative judgments is response time (RT), which contains potential information concerning respondents' decision-making process. It would be challenging but exciting to combine RT into FC measures better to reveal respondents' behaviors or preferences in personality measurement. Given this situation, this study aims to propose a new item response theory (IRT) model that incorporates RT into FC measures to improve personality assessment. Simulation studies show that the proposed model can effectively improve the estimation accuracy of personality traits with the ancillary information contained in RT. Also, an application on a real data set reveals that the proposed model estimates similar but different parameter values compared with the conventional Thurstonian IRT model. The RT information can explain these differences.

Scholars have long recognized that reading in digital spaces requires unique skills, strategies, and competencies in comparison to those needed for reading printed text. In recent years, the ubiquity of social media and algorithmically targeted content has radically changed the nature of online reading and meaning making. Technological changes have occurred simultaneously with radically altered sociocultural and sociopolitical contexts. To account for an altered technological and sociocultural landscape, new approaches to teaching digital reading and critical media literacy are needed. Addressing these concerns, this case study detailed a digital reading curriculum designed to be responsive to both the contemporary digital media environment and to students' out-of-school digital literacy practices and contexts. The curriculum was collaboratively designed by five middle-school language arts teachers who participated in a semester-long professional learning group focused on digital reading. Drawing upon sociocultural, asset-based, and culturally relevant philosophies of education, these five teachers designed a unique digital reading curriculum. This study examined the nature of this curriculum. The findings detailed four aspects of the teachers' unit: (1) digital reading instruction situated within students' literate lives; (2) critical instruction regarding systemic features of the internet such as algorithms and clickbait; (3) lessons in which students interrogate socially situated meaning making; and (4) lessons focused on the role of emotions while reading online. The findings have implications for future digital reading and media literacy curricula intended to be responsive to students' funds of knowledge, ever-changing literacy technologies, and new, emergent ways of reading and practicing literacy on the internet.

Learning outcomes can be predicted with machine learning algorithms that assess students' online behavior data. However, there have been few generalized predictive models for a large number of blended courses in different disciplines and in different cohorts. In this study, we examined learning outcomes in terms of learning data in all of the blended courses offered at a Chinese university and proposed a new classification method of blended courses, in which students were primarily clustered on the basis of their online learning behaviors in blended courses using the expectation-maximization algorithm. Then, the blended courses were classified on the basis of the cluster of students who were present in the course and had the highest proportion. The advantage of this method is that the criteria used for classification of the blended courses are clearly defined on the basis of students' online behavior data, so it can easily be used by machine learning systems to algorithmically classify blended courses based on log data collected from a learning management system. Drawing on the classification of the blended courses, we also proposed and validated a general model using the random forest algorithm to predict learning outcomes based on students' online behaviors in blended courses with different disciplines and different cohorts. The findings of this study indicated that after blended courses were classified on the basis of students' online behavior, prediction accuracy in each category increased. The overall accuracies for Course I (380 courses out of 661 after screening), L (14 courses out of 661 after screening), A (237 courses out of 661 after screening), V (8 courses out of 661 after screening), and H (22 courses out of 661 after screening) were 38.2%, 48.4%, 42.3%, 42.4%, and 74.7%, respectively. According to these results, it was found that a prerequisite for the accurate prediction of students' learning outcomes in a blended course was that most students should be highly engaged in a variety of online learning activities rather than being focused on only one type of activity, such as only watching online videos or submitting online assignments. The prediction model achieved accuracies of 80.6%, 85.3%, 63%, 54.8%, and 14.3% for grades A, B, C, D, and F in Course H, respectively. The results demonstrated the potential of the proposed model for accurately predicting learning outcomes in blended courses. Finally, we found that there was no single online learning behavior that had a dominant effect on the prediction of students' final grades.

Background: Life satisfaction is a key component of students' subjective well-being due to its impact on academic achievement and lifelong health. Although previous studies have investigated life satisfaction through different lenses, few of them employed machine learning (ML) approaches. Objective: Using ML algorithms, the current study predicts secondary students' life satisfaction from individual-level variables. Method: Two supervised ML models, random forest (RF) and k-nearest neighbours (KNN), were developed based on the UK data and the Japan data in PISA 2018. Results: Findings show that (1) both models yielded better performance on the UK data than on the Japanese data; (2) the RF model outperformed the KNN model in predicting students' life satisfaction; (3) meaning in life, student competition, teacher support, exposure to bullying and ICT resources at home and at school played important roles in predicting students' life satisfaction. Conclusions: Theoretically, this study highlights the multi-dimensional nature of life satisfaction and identifies several key predictors. Methodologically, this study is the first to use ML to explore the predictors of life satisfaction. Practically, it serves as a reference for improving secondary students' life satisfaction.

BRIDGES is a software framework for creating engaging assignments for required courses such as data structures and algorithms. It provides students with a simplified API that populates their own data structure implementations with live and real-world data, and provides the ability for students to easily visualize the data structures they create as part of routine classroom exercises. The objective is to use the infrastructure to promote a better understanding of the data structure and its underlying algorithms. This report describes the BRIDGES infrastructure and provides evaluation data collected over the first five years of the project. In the first 2 years, as we were developing the BRIDGES projects, our focus was on gathering data to assess whether the addition of the BRIDGES exercises had an effect on student retention of core concepts in data structures; and throughout the 5-year duration of the project, student interest and faculty feedback were collected online and anonymously. A mixed method design was used to evaluate the project impact. A quasiexperimental design compared student cohorts who were enrolled in comparable course sections that used BRIDGES with those that did not. Qualitative and quantitative measures were developed and used together with course grades and grade point averages. Interest and relevance in BRIDGES programming assignments was assessed with additional survey data from students and instructors. Results showed that students involved in BRIDGES projects demonstrated larger gains in knowledge of data structures compared to students enrolled in comparable course sections, as well as long-term benefits in their performance in four follow-on required courses. Survey responses indicated that some investment of time was needed to use BRIDGES, but the extra efforts were associated with several notable outcomes. Students and instructors had positive perceptions of the value of engaging in BRIDGES projects. BRIDGES can become a tool to get students more engaged in critical foundational courses, demonstrating relevance and context to today's computational challenges.

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.

More than 50 years ago, Bongard introduced 100 visual concept learning problems as a challenge for artificial vision systems. These problems are now known as Bongard problems. Although they are well known in cognitive science and artificial intelligence, only very little progress has been made toward building systems that can solve a substantial subset of them. In the system presented here, visual features are extracted through image processing and then translated into a symbolic visual vocabulary. We introduce a formal language that allows representing compositional visual concepts based on this vocabulary. Using this language and Bayesian inference, concepts can be induced from the examples that are provided in each problem. We find a reasonable agreement between the concepts with high posterior probability and the solutions formulated by Bongard himself for a subset of 35 problems. While this approach is far from solving Bongard problems like humans, it does considerably better than previous approaches. We discuss the issues we encountered while developing this system and their continuing relevance for understanding visual cognition. For instance, contrary to other concept learning problems, the examples are not random in Bongard problems; instead they are carefully chosen to ensure that the concept can be induced, and we found it helpful to take the resulting pragmatic constraints into account.

A shift in primacy from online participatory cultures to algorithmic cultures invites new questions about literacies in digital contexts. This article contributes to the conceptualisation of literacies in algorithmic cultures through sociomaterial and affect theories. It develops a sociomaterial perspective that proposes felt, observable moments of user-algorithmic co-productions of culture as a needed unit of analysis for researching contemporary, critical digital literacies. It then employs this unit as a starting point for analysing the interplay of feeling, critical reflection and algorithm agency across one young adult's self-described literacy practice of 'working algorithms' across social media platforms. Analysis illustrates how critical literacies in algorithmic cultures are driven by processes of human-machine feeling-thinking that cannot be reduced to rational critiques of ideologies, platform capitalism or other forms of power alone. It describes how Malaya became more attuned over time to the affects of working with platform algorithms to craft her community, her sense of self and her sense of well-being. This sensitivity to feeling moved and feeling the capacity to move machines through the use of her literacies highlights how the facilitation of affect is a crucial point of analysis in understanding contemporary digital literacies.

Algorithmic systems such as Learning Analytics (LA) are driving the datafication and algorithmization of education. In this research, we focus on the appropriateness of LA systems from the perspective of parents and students in secondary education. Anchored in the contextual integrity framework (Nissenbaum, "Washington Law Review, 79," 41, 2004), we conducted two survey studies (N[subscript students]=277, N[subscript parents]=1013) in Flanders to investigate how they evaluate the appropriateness of the data flows in LA systems, and how both populations differ in their evaluations. The results show that the most-used student-centered LA are perceived less appropriate than the less-used teacher-centered LA by both students and parents. The usage of personal characteristics in LA is perceived as least appropriate, in contrast to coarser class characteristics. Sharing insights of LA with institutions that are part of the traditional educational context, such as the school, is seen as the most appropriate, and more appropriate than sharing it with learning platforms or third parties (e.g., Big Tech). Overall, we found that parents evaluated the different elements of the dataflows embedded in LA as less appropriate than students. In the discussion, we argue that educational institutions should include the evaluation of both parents and students to further manage expectations and construct shared norms and practices when implementing LA in education.informa

Monitoring students' academic progress is vital for ensuring timely completion of their studies and supporting at-risk students. Educational Data Mining (EDM) utilizes machine learning and feature selection to gain insights into student performance. However, many feature selection algorithms lack performance forecasting systems, limiting their ability to predict future academic success accurately. To address this, we propose FeatureX, a hybrid approach aiming to select high-performing features that impact student quality and reduce dropout rates. FeatureX integrates filter-based and wrapper-based methods to identify relevant features for predicting student performance. This approach enhances educational experiences by optimizing resource allocation and support services. Additionally, the Confidence-Weighted Fusion Voting Classifier (CWFVC) Algorithm supplements feature selection with performance forecasting capabilities, improving accuracy by combining diverse machine learning classifiers. The research evaluates FeatureX using Decision Trees, Random Forests, Support Vector Machines, and Neural Networks. Performance metrics, including accuracy, precision, recall, and F1-score, measure FeatureX's effectiveness. Results show that FeatureX achieves the highest accuracy with a subset of ten features, effectively identifying influential predictors. The CWFVC Algorithm further enhances performance forecasting accuracy, enabling timely identification of at-risk students and reducing dropout rates to foster inclusive education. Our research addresses the demand for data-driven approaches in education, offering an innovative method for predicting student performance and enhancing educational outcomes for diverse students. FeatureX and the CWFVC Algorithm provide valuable tools for educators and administrators to optimize resources, tailor support services, and create a more inclusive learning environment. Leveraging EDM and performance forecasting, educational institutions can proactively support students and promote academic success, contributing to an equitable and effective educational system.

Literature screening is the process of identifying all relevant records from a pool of candidate paper records in systematic review, meta-analysis, and other research synthesis tasks. This process is time consuming, expensive, and prone to human error. Screening prioritization methods attempt to help reviewers identify most relevant records while only screening a proportion of candidate records with high priority. In previous studies, screening prioritization is often referred to as automatic literature screening or automatic literature identification. Numerous screening prioritization methods have been proposed in recent years. However, there is a lack of screening prioritization methods with reliable performance. Our objective is to develop a screening prioritization algorithm with reliable performance for practical use, for example, an algorithm that guarantees an 80% chance of identifying at least 80% of the relevant records. Based on a target-based method proposed in Cormack and Grossman, we propose a screening prioritization algorithm using sampling with replacement. The algorithm is a wrapper algorithm that can work with any current screening prioritization algorithm to guarantee the performance. We prove, with mathematics and probability theory, that the algorithm guarantees the performance. We also run numeric experiments to test the performance of our algorithm when applied in practice. The numeric experiment results show this algorithm achieve reliable performance under different circumstances. The proposed screening prioritization algorithm can be reliably used in real world research synthesis tasks.

The rapid development of artificial intelligence has driven the transformation of educational evaluation into big data-driven. This study used a systematic literature review method to analyzed 44 empirical research articles on the evaluation of big data education. Firstly, it has shown an increasing trend year by year, and is mainly published in thematic journals such as educational technology, science education, and language teaching. Chinese and American researchers have made the greatest contributions in this field. Secondly, the algorithmic models for big data education evaluation research are diverse, the text modality is the most popular, the evaluation subjects are mainly college students, with fewer primary and secondary school students, and science is the discipline that most commonly applies big data education evaluation. The evaluation objectives of big data education evaluation mainly focus on five aspects: high-order thinking analysis, learning performance prediction, learning emotion recognition, teaching management decision-making, and evaluation mode optimization, and the text modality is widely used for data collection in high-order thinking analysis; regardless of the evaluation objectives, higher education students are the most widely evaluated objects; the science discipline is the main field of using big data technology to empower teaching evaluation. Thirdly, the current research topics of big data education evaluation mainly focus on online learning behavior and environmental participation evaluation, process assessment of learning motivation and emotional analysis, development and optimization of subject domain big data models, cognitive diagnosis and high-order thinking skills evaluation, and design of learning analysis frameworks based on data mining.