The tool, called SALS-BioC, is described in the journal SoftwareX as a soft-sensor based adaptation learning simulator for predicting bacterial concentrations in membrane bioreactors. A soft sensor, in the engineering sense, is a machine learning model that estimates hard-to-measure quality variables from variables that are easy to measure continuously. In this case, the system takes routine physicochemical water quality readings such as pH, electrical conductivity, total suspended solids, biochemical oxygen demand, chemical oxygen demand, nitrate nitrogen, turbidity, and chlorine levels, and uses them to predict how many bacterial cells are present in the treated effluent. The idea is to replace days of laboratory work with an instant, data-driven estimate that plant operators can act on in real time.

The research team, composed of H. Bagci, I. N’Doye, F. Almulhim, and P.-Y. Hong, confronted a familiar obstacle in environmental machine learning: there is simply not enough data. Wastewater treatment plants face cost and confidentiality constraints that limit access to large datasets from water-based epidemiology studies, and microbial measurements are inherently scarce because they depend on weekly sampling campaigns. Machine learning models trained on small, static datasets tend to perform well in the laboratory but fail when confronted with new conditions, a problem known as poor generalization. When the researchers tested a conventional long short-term memory (LSTM) neural network on data from a biologically independent sampling period, the model achieved an impressive coefficient of determination of 0.96 on its own training and testing data, but collapsed to a negative value of minus 0.33 on the unseen replicate, meaning its predictions were worse than a naive average.

To solve the data scarcity problem, the team developed a novel synthetic data generation algorithm called EMCM-PS, short for an ensemble Markov chain model with a joint state representation and probabilistic sampling rule. The approach is a creative departure from mainstream generative techniques. Instead of using generative adversarial networks, which pit two neural networks against each other in a training game and often suffer from a failure mode known as mode collapse, EMCM-PS builds on Markov chains and Gaussian copulas. Each multivariate observation in the wastewater dataset is encoded as a single joint state, and new synthetic samples are drawn directly from the empirical probability distribution of those observed states rather than from a sequential transition matrix. The researchers argue this design choice matters because wastewater samples are collected at irregular intervals and do not form a strict time series, so forcing a random-walk structure onto the data would introduce spurious dependencies and trap the generator in a limited set of states.

The generation workflow begins by augmenting the original dataset with correlation-aware multiplicative noise, then discretizing the samples into quantile-based joint states. Synthetic states are sampled independently from the global frequency distribution of observed states and decoded back into continuous measurements using either a local mode, which samples from the observations associated with each state, or a uniform mode, which draws values within the bounds of the corresponding quantile bucket. The software includes a suite of visual diagnostics to verify that the synthetic data faithfully reproduce the statistical structure of the real measurements, including marginal distribution comparisons, Pearson correlation heatmaps, and dimensionality-reduction plots based on principal component analysis and t-distributed stochastic neighbor embedding. A fidelity scorecard summarizes these metrics so users can judge at a glance whether the generated data preserve the variability and dependency patterns of the original samples.

On top of this synthetic data engine sits the second key innovation: a lifelong learning framework built around the LSTM predictor. Unlike traditional machine learning models that are trained once on a static historical dataset and then frozen, the lifelong learning approach continuously updates the model as new batches of data arrive from the target environment, while preserving previously acquired knowledge through a dictionary learning mechanism. In the SALS-BioC workflow, a source-domain model is first trained on historical data, then adapted to a target-domain dataset processed in sequential batches of thirty samples. For each batch, the model predicts first and is updated with the true values only afterward, mimicking the realistic operational scenario in which laboratory confirmation lags behind the need for a prediction. The software even provides an animated visualization of how the root mean square error evolves batch by batch during adaptation, showing how quickly the model recovers its accuracy as it encounters the new domain.

The validation experiment drew on real data from the KAUST aerobic membrane bioreactor wastewater treatment plant, which treats a mix of municipal wastewater. Two biologically independent replicates were used, one collected weekly from July to October 2023 for model development and another from February to March 2024 to assess generalization. Fourteen physicochemical parameters were measured alongside bacterial abundances quantified by flow cytometry. When the lifelong learning model, trained on EMCM-PS-generated synthetic data, was adapted to the unseen replicate, it achieved a cross-validation coefficient of determination of 0.8723. When the same experiment was run using synthetic data from a Wasserstein generative adversarial network, the best-performing member of the GAN family, the score reached only 0.6450. That difference of 0.2273 translates into a 35.2 percent improvement for the new probabilistic approach, a striking margin in a field where incremental gains are the norm.

The software itself is designed to be accessible rather than the exclusive province of machine learning specialists. It is implemented as a modular Python application with a browser-based interface built on Flask, HTML, CSS, and JavaScript, using standard scientific libraries including NumPy, pandas, SciPy, scikit-learn, and PyTorch. Users upload CSV datasets, select target variables, configure parameters, and run simulations through three dashboard sections covering synthetic data generation, LSTM-based prediction, and lifelong learning adaptation. Trained models can be downloaded as serialized PyTorch files and reloaded later for validation without retraining, and Bayesian hyperparameter optimization is built in with configurable search iterations. The code is released under the MIT license on GitHub, and the authors note that the architecture deliberately separates the interface, the server orchestration, and the algorithms, so that new generators, predictive models, or adaptation strategies can be added through dedicated API routes without redesigning the application. Extensions to viral contaminant prediction are described as a natural next step.

The implications reach beyond one treatment plant in Saudi Arabia. Water-based epidemiology has surged in prominence since the COVID-19 pandemic demonstrated that sewage can serve as an early-warning system for disease outbreaks in entire communities, but the field remains bottlenecked by the pace of laboratory analysis. A reliable soft sensor that predicts bacterial concentrations from measurements plants already collect could enable continuous biological monitoring, early diagnosis of operational faults, and rapid response to contamination events, all without waiting for culture results. The authors also emphasize the educational value of the tool, arguing that its guided interface lowers the expertise required to design data-driven soft sensor systems and makes the technology accessible to interdisciplinary researchers and students.

The researchers are candid about limitations. EMCM-PS currently uses a fixed number of quantile-based buckets for all features, but different water quality variables have different distributions and may require different discretization resolutions, particularly under operational variations such as fluctuations in flow rate or for highly skewed measurements like total cell counts. Future work, they suggest, could develop an adaptive discretization strategy that tunes the bucket count for each feature dynamically. They also plan to validate the simulator across international wastewater datasets to establish its transferability. For now, SALS-BioC stands as a concrete demonstration that thoughtful statistical modeling, rather than ever-larger neural networks, can sometimes deliver the biggest wins when data are scarce, and that open, reproducible software may be the fastest route to putting adaptive artificial intelligence into the pipes and pumps of the world’s water infrastructure.

Subject of Research: Machine learning-based soft sensor prediction of bacterial concentrations in membrane bioreactor wastewater treatment

Article Title: SALS-BioC: A soft-sensor based adaptation learning simulator for predicting bacterial concentrations in membrane bioreactors

Article References: Bagci, H., N’Doye, I., Almulhim, F., & Hong, P.-Y. (2026). SALS-BioC: A soft-sensor based adaptation learning simulator for predicting bacterial concentrations in membrane bioreactors. SoftwareX, 36, Article 103084. https://doi.org/10.1016/j.softx.2026.103084

Image Credits: AI Generated

DOI: 10.1016/j.softx.2026.103084

Keywords: soft sensors, machine learning, wastewater treatment, membrane bioreactors, synthetic data generation, Markov chains, LSTM networks, lifelong learning, water-based epidemiology, bacterial concentration prediction, generative adversarial networks, environmental monitoring

Most existing digital literacy scales trace their lineage to the European Commission’s Digital Competence Framework for Citizens, known as DigComp. The framework, first published in 2013 and most recently updated as DigComp 2.2 in 2022, breaks digital competence into five areas covering information and data literacy, communication and collaboration, digital content creation, safety, and problem solving. The researchers argue that while these frameworks are comprehensive in many respects, they underweight the cognitive machinery that underpins effective technology use. Computational thinking, a concept popularized by computer scientist Jeannette Wing in 2006, encompasses decomposition, pattern recognition, abstraction, and algorithm design, and is increasingly viewed as a foundational literacy rather than a specialist programming skill.

The omission matters, the authors contend, because computational thinking and lifelong learning are not merely academic abstractions. They are directly linked to an individual’s prospects in the digital economy and, by extension, to overall quality of life, the central concern of the journal in which the study appears. Prior research has connected digital skills to employment quality and life satisfaction, and international bodies including the OECD and the World Bank have warned that the changing nature of work will reward those who can adapt, learn continuously, and reason systematically with digital tools. A measurement scale that ignores these capacities, the team reasoned, offers an incomplete picture of whether students are genuinely prepared for the world they will graduate into.

To construct the new instrument, the researchers combined the updated DigComp 2.2 framework with China’s own digital literacy skills and certification standards, developed by the Computer Education Research Association of Chinese Universities. This dual grounding reflects a deliberate effort to blend an internationally recognized competency model with nationally relevant policy requirements, since China has made digital literacy promotion a formal policy priority through action outlines issued by the Central Cyberspace Affairs Commission in 2021 and subsequent annual work plans. The resulting scale organizes digital literacy into four dimensions: digital knowledge and skills, computational thinking, digital social responsibility, and digital learning and collaborative innovation.

The psychometric evaluation drew on responses from 970 first-year students at a Chinese application-oriented undergraduate university, an institution type that emphasizes practical, career-focused education. The team subjected the data to rigorous statistical testing, examining reliability and validity across the four hypothesized dimensions. The analysis, which employed structural equation modeling techniques appropriate for categorical response data, demonstrated strong reliability and validity, confirming that the 18 items hang together coherently within their intended factors and that the four-dimension structure fits the observed data. The researchers also reported evidence of convergent and discriminant validity following established criteria in the measurement literature, indicating that the scale measures what it claims to measure and distinguishes digital literacy from related but distinct constructs.

With the instrument validated, the team turned to the substantive question of who scores well and where the gaps lie. On gender, the findings were nuanced. Overall digital literacy showed no significant difference between male and female students, but the dimension-level results diverged in telling ways: male students outperformed female students in computational thinking, while female students excelled in digital social responsibility, the dimension covering ethical conduct, safety, and responsible participation in digital society. These contrasting profiles suggest that aggregate scores can mask meaningful differences in the specific competencies that compose digital literacy, and that interventions targeting computational thinking may need to address gendered patterns in confidence, exposure, or encouragement.

Regional disparities emerged as a second major finding. Students from urban backgrounds demonstrated significantly higher overall digital literacy than their rural counterparts, a result consistent with earlier Chinese studies that have documented gaps in digital media literacy between urban and rural primary school students. Notably, however, computational thinking showed no significant regional differences, suggesting that the cognitive dimension of digital literacy may be more evenly distributed than access-dependent skills such as device familiarity, software fluency, or online information evaluation. The researchers also found that only-child students exhibited significantly higher digital literacy than students with siblings, while computational thinking again showed no significant difference between the two groups, a pattern that may reflect differences in household resources and parental attention rather than underlying cognitive capacity.

The scale’s designers emphasize that the instrument is intended for practical deployment, not just academic description. Universities can use it to diagnose entering students’ strengths and weaknesses, guide targeted training programs, and optimize digital education strategies across curricula. Because the scale is compact at 18 items, it can be administered efficiently to large cohorts, making it feasible for institutions to track changes over time or evaluate the effectiveness of specific interventions. Beyond the campus, the researchers suggest the tool enables scholars to examine digital literacy disparities across regions, institutions, and academic disciplines, and to explore how digital literacy correlates with outcomes such as academic satisfaction, employability confidence, and broader quality-of-life indicators.

The study arrives at a moment when measurement frameworks themselves are evolving. A fifth edition of the European framework, DigComp 3.0, has been published, and the OECD together with the European Commission released an AI literacy framework for primary and secondary education in 2026, underscoring how quickly the competency landscape is shifting. The Chinese team’s decision to integrate computational thinking anticipates this trajectory, treating the ability to reason algorithmically and learn continuously as core components of digital citizenship rather than optional extensions. As generative artificial intelligence tools spread through higher education, the boundary between using technology and understanding it is becoming a critical dividing line, and instruments that can locate students on that boundary are likely to grow in importance.

The research also carries implications for equity. If urban students systematically outperform rural students in overall digital literacy while computational thinking remains evenly distributed, then policy interventions may be most effective when they target resource-dependent skills, such as access to devices, connectivity, and structured digital instruction, rather than assuming fixed cognitive differences. Similarly, the gendered pattern across dimensions points to the value of dimension-specific diagnostics over single composite scores. The authors note that all relevant data generated or analyzed during the study are available from the corresponding author on reasonable request, and the work was funded by a Xiamen City educational science research project focused on digital literacy measurement in application-oriented undergraduate universities. For educators and policymakers wrestling with how to prepare students for a digital decade, the study offers both a sharper lens and a reminder that what we choose to measure shapes what we choose to teach.

Subject of Research: Development and psychometric validation of a digital literacy scale for college students integrating computational thinking and lifelong learning

Article Title: Developing and Validating a Digital Literacy Scale Integrating Computational Thinking and Lifelong Learning for College Students

Article References: Guo, L., Zhang, M., Hu, Y., Wang, N., Yang, J., Li, C., & Zhu, L. (2026). Developing and Validating a Digital Literacy Scale Integrating Computational Thinking and Lifelong Learning for College Students. Social Indicators Research, 184(3), Article 55. https://doi.org/10.1007/s11205-026-03931-8

Image Credits: AI Generated

DOI: 10.1007/s11205-026-03931-8

Keywords: digital literacy, computational thinking, lifelong learning, scale development, psychometrics, college students, DigComp 2.2, digital social responsibility, urban-rural gap, gender differences, digital education, quality of life

A credit transfer system is, at its core, an accounting mechanism for learning. It records verified educational achievements so that a course completed at one institution can be recognized at another, supporting lifelong learning in an economy where skills rapidly become obsolete. Similar systems have been tried internationally, from the European Credit Transfer and Accumulation System to South Korea’s Academic Credit Bank and proposals for aligned Asian academic credits. Yet the Chinese version is distinctive in scale and in the complexity of its governance: it must coordinate individual learners deciding whether to participate, schools deciding how much to invest in compatible teaching resources, and a government deciding how aggressively to supervise and subsidize the enterprise. Each actor’s payoff depends on what the others do, which is precisely the situation game theory was built to describe.

He and Bu constructed a three-subject dynamic evolutionary game model. In an evolutionary game, agents are not treated as perfectly rational calculators who instantly find the best move; instead, populations of players adjust their strategies over time, with more successful strategies spreading through imitation and reinforcement. The approach, which traces back to Lewontin’s application of game theory to evolutionary biology and has since become standard in economics, is well suited to education policy, where participants learn, imitate, and adapt rather than optimize from the start. The researchers incorporated external constraints, strategic assumptions, and payment assumptions into the model, grounding the payoff structure in the benefit relationships among the stakeholders.

The model assigns each player a binary strategic choice. Learners can adopt strong participation, actively engaging with the system to improve their abilities, or hold back. Schools can participate actively by building high-quality teaching resources that feed the credit system, or invest minimally. The government can exert strong dominance through supervision and support, or take a laxer stance. Each choice carries costs and expected benefits: learners pay tuition and effort in exchange for portable credentials and improved skills; schools bear the expense of developing quality courses and administrative compatibility in exchange for enrollment, reputation, and possibly government incentives; the government funds oversight and subsidies in exchange for social returns such as a better-trained workforce.

To move beyond abstract equilibrium conditions, the authors ran numerical simulations of the model. Simulations of this kind allow researchers to trace how the probability of each strategy changes over successive rounds, revealing the decision-making mechanism of the three parties under different parameter settings. The exercise is particularly valuable because the equilibria of evolutionary games depend sensitively on cost allocations and benefit magnitudes; a system that looks stable on paper can collapse if, for example, the cost of building compatible teaching resources falls entirely on schools while the benefits accrue mostly to learners and the state.

The central finding is a matching rule. The optimal configuration for constructing a suitable credit transfer system combines strong participation by learners focused on ability improvement, active participation by schools in creating high-quality teaching resources, and strong dominance by the government in supervision and support. In other words, no single actor can carry the system: learner enthusiasm without quality courses produces credentials with little content, and quality resources without learners generate waste, while absent government supervision undermines trust in the credits themselves. The system functions as a three-legged stool, and the evolutionary dynamics determine whether all three legs stabilize simultaneously.

The analysis also illuminates why credit transfer efforts elsewhere have struggled. International experience, from pan-European grading scales to attempts at equivalency between the European credit systems for higher and vocational education, shows that technical compatibility is only part of the problem; incentives matter just as much. The Chinese study suggests that misalignment of cost and benefit among stakeholders, rather than flaws in credit accounting itself, may explain stalled adoption. When the government’s supervision and support are strong enough to shift expected payoffs, the simulations indicate that learners and schools can be tipped into the cooperative strategies that make the whole arrangement viable.

Methodologically, the paper illustrates a growing trend in education research: importing formal tools from economics and biology to model policy systems with many interacting agents. The authors draw on stakeholder theory, which originated in management science with Donaldson and Preston’s influential formulation, to justify the three-party structure, and then embed it in a replicator-style dynamic framework. Numerical simulation serves as the bridge between analytic equilibrium conditions and practical policy levers, letting the researchers ask counterfactual questions about what happens when costs are reallocated or benefit demands change.

For policymakers, the practical implications are concrete. Investment in high-quality teaching resources should be encouraged so that schools find active participation genuinely rewarding, and government supervision should be robust enough to certify credit quality, since the value of any transferable credit rests on trust in its standards. Meanwhile, the design should make ability improvement, not merely credential collection, the visible payoff for learners, aligning individual motivation with the system’s educational purpose. The work was supported by China’s National Program for Funding Postdoctoral Researchers, and both authors are affiliated with Beijing Normal University.

As lifelong learning becomes an economic necessity in an era of rapid technological change, credit transfer systems are likely to spread globally, and the Chinese experience will be watched closely. He and Bu’s evolutionary game model provides a template for anticipating how such systems behave before they are built: identify the players, quantify their costs and benefits, simulate the dynamics, and design interventions that steer all parties toward the cooperative equilibrium. The finding that strong learner participation, active school engagement, and strong government dominance must arrive together is a warning against piecemeal reform, and a demonstration that the mathematics of strategic interaction can inform one of education’s most intricate institutional projects.

Subject of Research: Game-theoretic modeling of stakeholder participation strategies in China's credit transfer system for lifelong learning

Article Title: Participating Strategy of the Constructors in the Construction of Credit Transfer System in China

Article References: Participating Strategy of the Constructors in the Construction of Credit Transfer System in China. (n.d.). https://doi.org/10.1007/s44366-026-0077-z

Image Credits: AI Generated

DOI: 10.1007/s44366-026-0077-z

Keywords: credit transfer system, evolutionary game theory, lifelong learning, China, stakeholder theory, education policy, learners, schools, government supervision, numerical simulation, Beijing Normal University, Participating

The first of these domains, the integration of AI into educational systems, is no longer speculative. Machine learning techniques now power adaptive learning platforms, intelligent tutoring systems, automated feedback engines, and generative tools that can draft essays, solve problems, and answer questions in natural language. A systematic review published in Discover Education in 2025 catalogued the techniques used for personalized learning in adaptive education, documenting how recommendation algorithms and learner models adjust the pace, sequence, and difficulty of material for individual students. Related work in AI Magazine has described the underlying architecture of adaptive learning technologies, which typically combine statistical models of student knowledge with content repositories and decision engines that select the next best activity. The technical promise is considerable: a well-designed tutoring system can diagnose misconceptions in real time and intervene precisely where a learner struggles, something even the most dedicated teacher cannot do for thirty students simultaneously.

Yet the evidence also shows that personalization is not a solved problem. A 2024 preprint by researchers including Clément, Sauzéon, Roy, and Oudeyer reported improved performance and motivation in intelligent tutoring systems when machine learning recommendations were combined with learner choice, suggesting that autonomy matters as much as algorithmic accuracy. The lesson, the new paper suggests, is that personalization should augment human agency rather than replace it. Systems that hand students a menu of options alongside data-driven recommendations tend to produce better motivational outcomes than systems that silently steer learners along a single optimized path. This finding carries weight for developers and procurement officers alike, because the design choices embedded in commercial platforms often determine whether students experience AI as a tool they control or a system that controls them.

The second domain, learner development and personalization, raises the question of what students actually need to know in a world saturated with intelligent machines. A growing body of scholarship argues that AI literacy must become a foundational competency, on a par with reading, writing, and numeracy. A review in Interactive Learning Environments examined AI literacy education in secondary schools and found a rapidly expanding but still uneven landscape of curricula, teaching materials, and assessment approaches. In May 2025, the OECD and the European Commission jointly launched an AI literacy framework for primary and secondary education, an attempt to define what young learners should understand about how AI systems work, what they can and cannot do, and how they shape the information environment. Complementing this, the AI4AL project has focused on adult learning, recognizing that lifelong learners also need the capacity to evaluate and interact with AI tools in the workplace and in civic life.

Curriculum and educational content form the third domain, and here the pressures are acute. Research published in Education and Information Technologies in 2025 explored the impact of AI on curriculum development in higher education institutions worldwide, documenting how universities are redesigning programs to include machine learning, data ethics, and human-computer interaction, while simultaneously grappling with how generative AI undermines traditional take-home assessments. Some institutions have gone further and created dedicated degrees; the Autonomous University of Barcelona, for example, now offers a bachelor’s degree in artificial intelligence. Beyond specialist programs, scholars have examined how AI can support interdisciplinary learning, with a systemic review in Education and Information Technologies synthesizing the application contexts, roles, and influences of AI across disciplines. Co-design approaches have also emerged: researchers at the Thirty-Eighth AAAI Conference on Artificial Intelligence described how cross-disciplinary high school teachers worked with computer scientists to build AI curricula that fit local classrooms, an approach the new paper presents as a model for content development that is both technically sound and pedagogically grounded.

The fourth domain concerns the role of teachers, arguably the most contested question in the entire debate. The paper argues against narratives of replacement and for a redefinition of professional practice. Teachers bring something that current AI systems do not: expectations. A narrative review by Rubie-Davies and Hattie published in the Journal of the Royal Society of New Zealand synthesized decades of evidence showing that teacher expectations exert a powerful influence on student achievement. An algorithm that ranks students by predicted performance, if absorbed uncritically by teachers, could amplify bias; used reflectively, the same data could help teachers question their own assumptions. The literature also points to creative applications, surveyed in a 2024 book by Urmeneta and Romero, in which AI supports artistic and design-oriented learning, and to responsible AI education, examined by Aler Tubella, Mora-Cantallops, and Nieves in Ethics and Information Technology, which prepares students to reason about the ethical dimensions of the systems they build and use. Teacher training for these new roles remains a bottleneck, and the paper joins a broader chorus in calling for sustained professional development rather than one-off workshops.

Assessment deserves particular attention because generative AI has destabilized the essay, the problem set, and the take-home exam in a matter of months. A 2024 paper in the International Journal of Educational Technology in Higher Education argued that the classroom response must combine AI literacy, prompt engineering, and critical thinking, teaching students to work with generative tools while evaluating their outputs. Chasokela and Hlongwane, writing on assessing higher-order and critical skills in the era of artificial intelligence, explore how examinations can be redesigned to measure judgment, synthesis, and originality rather than reproduction of information that a language model can generate in seconds. The technical challenge is real: detectors of AI-generated text are unreliable, and arms races between generation and detection technologies offer little assurance. The more promising route, the paper suggests, is to change what is assessed and how, embedding tasks in live performances, oral defenses, project portfolios, and processes that make reasoning visible.

The fifth domain, governance and regulation, frames all the others. The European Union’s Regulation 2024/1689, the so-called AI Act, establishes harmonized rules for artificial intelligence across sectors, including requirements relevant to educational applications. UNESCO has issued both a Recommendation on the ethics of artificial intelligence and a practical guidance document for policy-makers on AI and education, urging governments to adopt human-centered approaches and to protect the rights of learners. Schiff’s influential 2022 article in the International Journal of Artificial Intelligence in Education made the case for education for AI, not merely AI for education, arguing that national AI strategies systematically undervalue the role of education and ethics in preparing citizens. Hu and colleagues, surveying AI in higher education, document how policy development lags behind deployment, leaving institutions improvising rules for tools they have already adopted. The new paper calls for collaborative policymaking that includes teachers, students, families, and researchers, not only technology vendors and ministries.

What emerges from the synthesis is a call for systemic thinking. The five domains are interdependent: personalization technologies shape curricula, curricula redefine teacher roles, teacher roles demand new assessments, and all of it requires governance that anticipates rather than reacts. The author argues that ethically grounded approaches must promote equity, ensuring that AI-rich benefits do not accrue only to well-resourced schools while underfunded systems receive automation as a substitute for investment. Lifelong learning is a second pillar, because the pace of technological change means that a single front-loaded education can no longer carry a person through a career. Collaborative policymaking is the third, distributing authority so that decisions about classroom AI are not left to procurement contracts alone. The work was supported by the Ministry of Science and Innovation of Spain, and the author acknowledges the GDEI Research Team of the Chinese Academy of Educational Sciences and feedback from Margarida Romero. As a brief communication, the paper does not present new empirical data; its contribution is a structured map of a fast-moving field and an argument about direction. That argument is likely to resonate widely, because it refuses both utopian and catastrophist framings and insists instead that the outcome depends on choices societies make now about curriculum, pedagogy, and law. In classrooms from Barcelona to Beijing, those choices are already being made, one lesson plan and one procurement decision at a time.

Subject of Research: Integrating artificial intelligence into education systems across personalization, curriculum, teaching, assessment, and governance.

Article Title: Educational Transformation in the Era of Artificial Intelligence

Article References: Manyà, F. (2026). Educational Transformation in the Era of Artificial Intelligence. Frontiers of Digital Education, 3(1), Article 8. https://doi.org/10.1007/s44366-026-0082-2

Image Credits: AI Generated

DOI: 10.1007/s44366-026-0082-2

Keywords: artificial intelligence in education, personalized learning, AI literacy, intelligent tutoring systems, education policy, teacher roles, curriculum development, assessment, ethics, governance, lifelong learning, generative AI