Abstract This article examines the challenges primary and secondary teachers face in implementing formative assessment, with a particular focus on the use of digital technologies and the emerging potential of artificial intelligence (AI), including generative AI (GenAI) and agentic AI. Drawing on empirical research and theoretical perspectives, we explore how formative assessment—an established pedagogical practice with significant impact on student learning—has evolved alongside technological developments. We revisit a well‐established framework of five key formative assessment strategies, analysing how it has been extended to integrate digital technologies and how this influences the roles of teachers, learners and tools in classroom decision‐making. Our central research question asks: How can GenAI be integrated into formative assessment practices to enhance student learning while supporting teacher agency and professional judgement? We argue that GenAI, when critically and thoughtfully deployed, can create new opportunities for personalised feedback, dynamic learning pathways and the co‐construction of knowledge between teachers and students. Moreover, GenAI's ability to support ‘moments of contingency’ enables teachers to respond more effectively to emerging learning needs, thus fostering self‐regulation and deeper engagement. However, we stress that AI agency is agency without intelligence. The value of these technologies depends on how they are interpreted and implemented by educators, which requires ongoing reflection, collaboration and theoretical understanding. With deep implications for teacher education programmes, our analysis suggests that teacher quality in this evolving pedagogical landscape should be understood as adaptive, multifaceted and grounded in both technological fluency and sound formative assessment principles, moving beyond the narrow and prescriptive definitions that dominate recent educational policy in England. Practitioner notes What is already known about this topic Formative assessment is well established in education, with strong empirical support showing high impact on learning. Digital technologies have shown more variable and generally moderate impact on learning, despite long‐standing interest. Teachers often face challenges implementing formative assessment effectively due to limited training and systemic constraints. Very little is known about how generative AI (GenAI) will influence formative assessment practices, given its novelty and potential to disrupt traditional teaching roles. What this paper adds This paper examines how formative assessment can be enhanced—and complicated—by digital technologies, especially new forms of AI. A formative assessment framework is reviewed, outlining how GenAI and agentic AI could support each of its five core strategies. The paper offers a timely analysis of AI's pedagogical potential, portraying it as a dynamic but fallible agent in learning. Implications for practice and/or policy Teacher education must now include AI literacy alongside formative assessment pedagogy, enabling ethical and informed use of GenAI. Educators should be equipped to evaluate and adapt emerging technologies while protecting student agency and pedagogical intent. Schools and policymakers must foster conditions for early‐career teachers to explore and refine AI‐supported approaches. Ethical safeguards are needed to address concerns around feedback quality, learner autonomy and dependency on automation. Policy should promote collaborative design of pedagogies that integrate AI responsibly, centring on deep learning and teacher–student relationships.

This study engages with the ongoing theoretical debate between the technological solutionism prevalent in digital education (Morozov, 2013) and critical pedagogical perspectives that question the socio-technical implications of technology integration (Fenwick et al., 2015). In nursing education, this manifests as a tension between evidence-based advocacy for digital tools, such as instructional videos, and the imperative to foster pedagogical depth that cultivates critical thinking and clinical reasoning. However, scant research has systematically examined how the structured fusion of a robust pedagogical model such as Problem-Based Learning (PBL) with video-based scaffolding reconfigures the learning ecology, thereby affecting both procedural competency and learner agency. This article addresses this gap by asking: How does a systematically developed Video-Supported PBL model influence nursing students’ clinical skill acquisition and critical thinking? Using a design-based research (DBR) approach, the study developed, validated, and empirically tested the “MEDIFA” model, integrating expert validation and pretest/posttest analyses with 30 students. The findings demonstrate significant quantitative gains in clinical competency (N-Gain = 0.76) and reveal a qualitative shift toward self-regulated learning, in which students strategically used on-demand videos for mastery while engaging in collaborative problem-solving. The analysis further shows a reconfiguration of the instructor’s role from primary demonstrator to facilitator of reasoning, mediated by the digital scaffold. The article argues that this integration creates a synergistic learning system where cognitive load management via video enables deeper participation in situated, practice-based communities (Lave & Wenger, 1991), thereby bridging a key conceptual divide in the literature. It contributes a validated instructional model and a refined theoretical synthesis, offering a more nuanced framework for designing and evaluating technology-enhanced learning in competency-based professional education.

MsTargetPeaker: a quality-aware deep reinforcement learning approach for peak identification in targeted proteomics.

Examining the Quality and Quantity of Verbal Feedback in the Operating Room: A Multi-Specialty Study in a Canadian Context.

Cryptocurrency financial markets are characterized by high volatility and non-stationary price dynamics, posing significant challenges to traditional portfolio optimization techniques that rely on static risk–return assumptions. In such environments, existing methods often struggle to generalize and adapt effectively, leading to suboptimal performance and increased downside risk. To address these limitations, this paper proposes a novel adaptive portfolio optimization framework that integrates Generative Adversarial Networks (GANs) for synthetic data augmentation with a state-of-the-art Soft Actor-Critic (SAC) deep reinforcement learning (DRL) agent. By augmenting real historical OHLC data with realistic TimeGAN-generated price sequences, the proposed approach exposes the DRL agent to a broader range of market scenarios, thereby improving generalization and mitigating overfitting. A convolutional neural network (CNN) feature extractor captures deep temporal dependencies, while causal and dilated convolutions model complex inter-asset correlations. Empirical results demonstrate that the proposed GAN–SAC hybrid consistently outperforms conventional strategies and the baseline Deep Portfolio Optimization (DPO) model, achieving a higher Accumulative Portfolio Value (APV) of 53.72, an improved Sharpe Ratio of 0.0980, and a reduced Maximum Drawdown (MDD) of 28.5%. These findings confirm the effectiveness of combining generative models and DRL to develop robust, adaptive portfolio strategies capable of navigating highly volatile cryptocurrency markets, with practical implications for next-generation algorithmic trading systems requiring enhanced resilience and dynamic risk control.

Explainable reinforcement learning for glucose monitoring based on shapley value analysis.

This paper proposes using reinforcement learning (RL) agents enhanced with 3D vision-language models (VLMs) to enable sustainable smart homes. These agents perceive the 3D layout and objects of a household environment and learn to autonomously adjust systems (e.g., HVAC, lighting, appliances) to optimize energy use and resource management. We identify specific energysaving tasks (such as occupancy-driven thermostat control, efficient lighting and blind management) and resource-management tasks (like waste sorting assistance and water-use feedback) that such agents can perform. We review recent advances in RL and vision-language models, and outline a conceptual framework for embodied home agents. Through this synthesis, we demonstrate how RL-powered agents can significantly reduce domestic energy consumption and waste, thereby supporting eco-friendly lifestyles. We also discuss the potential environmental and economic benefits of these systems, as well as technical and social challenges to their adoption. The contribution of this work is in articulating “spatial adaptation” for sustainability: an RL-driven approach that transforms smart homes into proactive, learning environments for green living.

Deep reinforcement learning (DRL) has shown great potential in many fields due to its powerful decision-making ability. To enable agents to acquire sufficient generalization capabilities, domain randomization is applied during the initialization phase of training in parameterizable environments. However, due to the commonly adopted uniform random sampling strategy, the agent will obtain inefficient samples from suboptimal environments in the late training stage, which limits the enlargement of the agent's effective decision region. To address this issue, the environmental difficulty is defined first based on the long-term rewards of the agent during training. Subsequently, we proposed an adaptive environment generator (AEG) based on the Gaussian mixture model (GMM), which dynamically generates training environments with corresponding difficulty levels tailored to the agent's learning progression. The generator maintains a database of environmental parameters based on environmental difficulty, and fits a GMM with the data in the database. During the environment initialization stage in each training episode, the AEG probabilistically generates environmental parameters through sampling from either the GMM or a uniform random distribution, ensuring both appropriate difficulty and sufficient exploration capability. Simulation results demonstrate that AEG-based training expedites learning in the early phases while generating more challenging environments in the late stages. Comprehensive evaluations across multiple environments validate the general applicability of AEG and demonstrate that the resulting agent achieves a broader coverage of effective decision-making regions compared to baseline methods.

Purpose Generative AI (GenAI) in education brings renewed attention to learner autonomy – that is, whether learners can think and act independently. GenAI offers the promise of learning efficiency and personalization, while raising questions about its alignment with nurturing autonomous learners. In this paper, we present a theoretical framework to investigate the relationship between GenAI and learner autonomy, to guide the design of educational environments that are safe and autonomy-supporting. Design/methodology/approach Our paper explores the multifaceted nature of autonomy across the cognitive, philosophical, political and computing fields, connecting theories such as self-determination theory with reflections on machine autonomy. Leveraging Latour's Actor-Network Theory, our framework aims to elucidate how autonomy is distributed between human and non-human actors in educational environments. Findings Our main contribution is the process of “autonomy budgeting”, viewing autonomy as a resource that is allocated and traded off between an ensemble of actors. Autonomy budgeting works as a guiding conceptual tool for researchers, educators, curriculum designers and policymakers to assess and manage the autonomy trade-offs involved in integrating GenAI into educational environments. Research limitations/implications By re-centering the learner's agency and capacity for self-regulation, autonomy budgeting provides a way to conceptualize and operationalize autonomy within AI-mediated education, and to navigate the complex interplay between human and machine agency in education. Originality/value Our framework develops reflections on the socio-technical nature of educational processes, where technologies act as co-participants rather than neutral tools. Autonomy in education, becomes a multifaceted construct that spans (human) cognitive, epistemic and political domains, and must be considered vis-a-vis varying degrees of machine autonomy.

It can be argued that Second Language Acquisition (SLA) classes are more visual that at any other time in history. Textbooks, worksheets, and even lectures are now often full of illustrations. Illustrations help learners form connections between the target language and the learner’s experience which enhances acquisition and usage (Canning-Wilson, 1999). Sketching, which is quickly drawing simple images with thick lines in order to illustrate an idea or object, may be a valuable tool in the classroom for three reasons: 1) sketching activates the Reticular Activating System (RAS) in the brainstem which increases alertness and retention of information, 2) sketching increases learner agency through active participation, and 3) sketching accommodates a variety of learning styles (Castillo, 2007). This paper has four objectives. 1) Analyze the previously mentioned benefits of sketching. 2) Briefly explore Horsburgh’s Four Central Guidelines for Class Illustrations and how teachers can use them to better utilize sketching in the classroom. 3) Offer sketching activities that students can use to better learn vocabulary. 4) Share examples of both student and teacher sketches are well as student feedback.

Autonomous unmanned aerial vehicles (UAVs) offer cost-effective and flexible solutions for a wide range of real-world applications, particularly in hazardous and time-critical environments. Their ability to navigate autonomously, communicate rapidly, and avoid collisions makes UAVs well suited for emergency response scenarios. However, real-time path planning in dynamic and unpredictable environments remains a major challenge, especially in confined tunnel infrastructures where accidents may trigger fires, smoke propagation, debris, and rapid environmental changes. In such conditions, conventional preplanned or model-based navigation approaches often fail due to limited visibility, narrow passages, and the absence of reliable localization signals. To address these challenges, this work proposes an end-to-end emergency response framework for tunnel accidents based on Multi-Agent Reinforcement Learning (MARL). Each UAV operates as an independent learning agent using an Independent Q-Learning paradigm, enabling real-time decision-making under limited computational resources. To mitigate premature convergence and local optima during exploration, Grey Wolf Optimization (GWO) is integrated as a policy-guidance mechanism within the reinforcement learning (RL) framework. A customized reward function is designed to prioritize victim discovery, penalize unsafe behavior, and explicitly discourage redundant exploration among agents. The proposed approach is evaluated using a frontier-based exploration simulator under both single-agent and multi-agent settings with multiple goals. Extensive simulation results demonstrate that the proposed framework achieves faster goal discovery, improved map coverage, and reduced rescue time compared to state-of-the-art GWO-based exploration and random search algorithms. These results highlight the effectiveness of lightweight MARL-based coordination for autonomous UAV-assisted tunnel emergency response.

ABSTRACT This article examines how Technical and Vocational Education and Training (TVET) lecturers in a South African college conceptualise a humanising pedagogy within Work Integrated Learning (WIL) practices. Against a backdrop of high youth unemployment and persistent skills mismatches, the study considers how WIL might be reframed to support employability while fostering learner agency. Using a qualitative, critical design, five purposively selected lecturers drawn from the college’s five campuses participated in the study. Data were generated through participatory visual methods (collage-making and a modified Mmogo-method®) combined with semi-structured group discussions. Findings reveal that although lecturers seldom used the term ‘humanising pedagogy’, their practices often reflected its principles: learner-centredness, dignity, and responsiveness to students’ lived realities. The visual artefacts and narratives highlighted tacit and affective dimensions of pedagogy that interviews alone might miss. However, humanising practices were constrained by compliance-driven frameworks that emphasised logbooks, placements, and bureaucratic requirements. The study acknowledges that students’ perspectives were mediated through lecturer accounts, suggesting future research should include learner voices and direct observations. Overall, the paper argues that humanising WIL holds transformative potential by nurturing purpose, workplace readiness, and entrepreneurial imagination. It contributes to debates on vocational reform by offering context-responsive guidelines for socially inclusive WIL.

Egocentric representations of the environment have historically been relegated to being used only for simple forms of spatial behaviour such as stimulus-response learning. However, in the many cases that critical aspects of policies are best defined relative to the self, egocentric representations can be advantageous. Furthermore, there is evidence that forms of egocentric representation might exist in the wider hippocampal formation. Nevertheless, egocentric representations have yet to be fully incorporated as a component of modern navigational methods. Here we investigate egocentric successor representations (SRs) and their combination with allocentric representations. We build a reinforcement learning agent that combines an egocentric SR with a conventional allocentric SR to navigate complex 2D environments. We demonstrate that the agent learns generalisable egocentric and allocentric value functions which, even when only additively composed, allow it to learn policies efficiently and to adapt to new environments quickly. Our work shows the benefit for egocentric relational structure to be captured, as well as allocentric. We offer a new perspective on how cognitive maps could usefully be composed from multiple simple maps representing associations between state features defined in different reference frames.

Curling presents a challenging continuous-control problem in which shot outcomes depend on long-horizon interactions between complex physical dynamics, strategic intent, and opponent responses. Despite recent progress in applying reinforcement learning (RL) to games and sports, curling lacks a unified environment that jointly supports stable, rule-consistent simulation, structured state abstraction, and scalable agent training. To address this gap, we introduce a comprehensive learning framework for curling AI, consisting of a full-sized simulation environment, a task-aligned Markov decision process (MDP) formulation, and a two-phase training strategy designed for stable long-horizon optimization. First, we propose a novel MDP formulation that incorporates stone configuration, game context, and dynamic scoring factors, enabling an RL agent to reason simultaneously about physical feasibility and strategic desirability. Second, we present a two-phase curriculum learning procedure that significantly improves sample efficiency: Phase 1 trains the agent to master delivery mechanics by rewarding accurate placement around the tee line, while Phase 2 transitions to strategic learning with score-based rewards that encourage offensive and defensive planning. This staged training stabilizes policy learning and reduces the difficulty of direct exploration in the full curling action space. We integrate this MDP and training procedure into a unified Curling RL Framework, built upon a custom simulator designed for stability, reproducibility, and efficient RL training and a self-play mechanism tailored for strategic decision-making. Agent policies are optimized using Soft Actor–Critic (SAC), an entropy-regularized off-policy algorithm designed for continuous control. As a case study, we compare the learned agent’s shot patterns with elite match records from the men’s division of the Le Gruyère AOP European Curling Championships 2023, using 6512 extracted shot images. Experimental results demonstrate that the proposed framework learns diverse, human-like curling shots and outperforms ablated variants across both learning curves and head-to-head evaluations. Beyond curling, our framework provides a principled template for developing RL agents in physics-driven, strategy-intensive sports environments.

This study proposes a smart energy management framework for a university campus microgrid aimed at reducing dependence on the main power grid and increasing the utilization of photovoltaic (PV) generation under dynamic load and environmental conditions. The core contribution is a two-stage approach that combines a genetic algorithm (GA) for static day-ahead optimization with a soft actor-critic (SAC) reinforcement learning (RL) agent performing adaptive supervisory management of microgrid active and reactive power flows via battery control. The GA provides an optimal reference schedule under forecasted conditions, while the SAC agent is trained on eight representative scenarios derived from measured PV generation and campus load data to adapt battery operation and grid exchange under uncertainty. The results show that the benefit of RL does not lie in reproducing the static GA solution, but in learning economically rationally adaptive behavior. In particular, the SAC agent exploits low-tariff periods and hedges against adverse PV conditions by proactively adjusting battery charging strategies in real time. This adaptive behavior addresses a key limitation of static optimization, which cannot respond to deviations from forecasted operation, and represents the main added value of the proposed framework. From a practical perspective, the GA-SAC architecture operates at a supervisory level with low computational requirements, making it suitable for scalable deployment in smart campus and smart city energy management systems.

This paper will discuss the meaning of culturally responsive teaching (CRT) as its application in digitally mediated classes. The analysis is based on a narrative review and thematic synthesis of theoretical resources, empirical works, teacher experience, and digital classroom content to address how cultural representation, linguistic diversity, and learner agency are negotiated in online, hybrid, and classroom settings. The evidence demonstrates that multimodal digital resources, multilingual participation plans, and student-centered evaluation broaden the potential of culturally sustaining pedagogy, whereas the disparity in access, platform structures and algorithmic mediation limits the implementation. This paper contends that digital CRT needs concerted pedagogical, institutional and technological responses to ensure equity and inclusion. The conclusion outlines the potential research directions in the future, such as comparative cross-cultural research, participatory research with marginalized communities, and design research on multicultural digital pedagogy. The study is part of the growing controversies on multicultural learning and digital equity in the increasingly globalized learning environment.

This paper discusses an on-device artificial intelligence (AI) solution for real-time, privacy-preserving fraud detection in smart financial environments, ensuring privacy-preserving consumer transactions. We suggest a distributed, on-device fraud detection solution that uses federated learning (FL) to improve privacy while detecting fraudulent transactions efficiently across decentralized smart environments. In this work, we used several models, including reinforcement learning (RL) agent and Random Forest, and we tested their performance using several measures like accuracy, precision, recall, and F-score, ensuring their applicability to smart environments with resource constraints. The recommended mechanism also uses t-Distributed Stochastic Neighbor Embedding (t-SNE) and Principal Component Analysis (PCA) to reduce dimensions of data, visualize the results, and evaluate the success rate of transactions classified as fraudulent and non-fraudulent. In our methodology, we applied data collection, data preprocessing, and cleaning, and we evaluated the metrics of selected models to allocate resources effectively and support decision-making processes in edge-based fraud detection systems within smart environments.

Efficient route planning technology is the core support for ensuring the successful execution of unmanned aerial vehicle (UAV) flight missions. In this paper, the coordination issue of global route planning and local real-time obstacle avoidance in complex mountainous environments is studied. To deal with this issue, a hierarchical route planning framework is designed, including global route planning and AI-based local route re-planning using deep reinforcement learning, exhibiting both flexible versatility and practical coordination and deployment efficiency. Throughout the entire flight, the local route re-planning task triggered by dynamic threats can be executed in real time. Meanwhile, a multi-model DQN (MMDQN) agent with a Monte Carlo traversal iterative learning (MCTIL) strategy is designed for local route re-planning. Compared to existing methods, this agent can be directly used to generate local obstacle avoidance routes in various scenarios at any time during the flight, which simplifies the complicated structure and training process of conventional deep reinforcement learning (DRL) agents in dynamic, complex environments. Using the framework structure and MMDQN agent for local route re-planning ensures the safety and efficiency of the mission, as well as local obstacle avoidance during global flights. These performances are verified through simulations based on actual terrain data.

ABSTRACT Formal coach education provision, particularly those courses targeted at participation and grassroots coaches, often resembles a linear, time-bound pathway, somewhat devoid of choice for coaches who are typically presented with a predetermined curriculum. To advance beyond such approaches, this conceptual paper presents our ‘Selection-Box’ (S-B) metaphor. Rooted in our personal experiences, empirical evidence, and Bernsteinian theory (i.e. classification and framing), the S-B seeks to depict the image of an excited coach, eagerly choosing what coach education provision to access, when, and in what order. Through this metaphor, we seek to prompt a discussion and potentially a reorientation of formal coach education design. Specifically, the S-B metaphor encourages policy makers and course designers to imagine and potentially move beyond a primary focus on content or assessment and (re)orientate towards a focus on the process of learning within the coach’s biographical, temporal, and spatial contexts. To operationalise the metaphor, the paper explains how choice, time, and knowledge may facilitate greater learner agency such that coaches can begin to select, sequence, and pace learning that they deem relevant to them and their coaching context. To illustrate, we cautiously provide examples not as prescriptions, but as further aids to policy makers and course designers. In doing so, the paper moves beyond deconstruction and towards reconstruction by (1) challenging existing content or assessment-led coach education to move towards a more process-focused approach to coach learning; (2) highlighting theoretically informed ways that policy and course design could empower coaches to frame and classify material; and (3) providing a discursive tool (i.e. the S-B metaphor) to prompt internal and external discussions and action within policy makers’ and course designers’ context.

The vast majority of work in computer vision focuses on developing and applying new machine learning models, which often rely on large amounts of labeled training data. However, annotation is costly and time-consuming. This paper presents an approach based on Deep Reinforcement Learning (DRL) to automatically generate new annotations and reduce human effort in the preparation of training data for supervised learning models in object detection. Our methodology introduces a virtual agent trained with human guidance, inspired by constructivist teaching methods, where interaction with a human teacher supports the agent's learning process. The proposed approach, named "Try a Little More" (TLM), employs active learning to identify uncertain cases and request human intervention during training, progressively enhancing the agent's ability to annotate autonomously. We evaluated our methodology on a mammography dataset, where the agent created bounding box annotations later used in a state-of-the-art supervised object detection algorithm. The results demonstrate that the proposed approach improves the quality and quantity of training data, surpassing existing reinforcement learning and human-computer interaction methods. Quantitatively, TLM generated 414 new annotations with an IoU of 0.86 and F1-score of 0.92, while increasing the mAP of a YOLO detector from 0.52 to 0.91, representing a 75% improvement in detection performance with 35% less human intervention. This contribution advances the field of Data-Centric AI by introducing a teaching-inspired methodology that combines human advice and reinforcement learning to accelerate the creation of annotations in domains with scarce labeled data.