Feedback is essential for learning, but its effectiveness relies heavily on how well it engages students in the educational process. Generative AI offers novel opportunities to efficiently produce rich, formative feedback, ranging from direct explanations to incrementally sequenced scaffolding designed to promote learner autonomy. Despite these capabilities, it is still unclear whether sequenced (layered) AI feedback – which provides encouragement and hints before revealing the correct answer – genuinely enhances engagement and learning outcomes. To investigate this, we randomly assigned 199 participants to receive either sequenced or non-sequenced AI-generated feedback. We evaluated its impact on learning performance, cognitive and behavioral engagement, and affective perceptions to understand how these factors mediate overall learning outcomes. Results show that sequenced feedback elicited slightly higher behavioral engagement and, as anticipated, was perceived as more encouraging and supportive of student independence. Concurrently, however, it induced a higher level of mental effort. Mediation analyses identified a positive affective pathway driven by perceived encouragement, which was completely counteracted by a negative behavioral pathway associated with the average number of tasks requiring three or more submissions; the cognitive pathway (mental effort) remained non-significant. Overall, sequenced feedback led to significantly poorer learning outcomes when compared to direct, non-sequenced feedback. These findings highlight a crucial trade-off: although sequenced AI scaffolding boosts engagement and positive user perceptions, it can have a detrimental effect on actual learning performance. By integrating analyses of outcomes, perceptions, and underlying mechanisms, this study provides nuanced insights for designing automated, AI-driven feedback systems.

The design of learning tasks represents a fundamental pedagogical process that directly shapes educational effectiveness, but this process is time-consuming and knowledge-intensive for teachers. The rapid emergence of generative artificial intelligence (GenAI) technologies has created unprecedented opportunities for supporting teachers' teaching practice. However, despite growing interest in teacher-GenAI collaboration, current research primarily focuses on preservice teachers and overall lesson plan design, with limited attention to challenges inherent in teacher-GenAI codesign of learning tasks or what teachers learn from such collaboration. To address this gap, we employed a mixed-methods approach involving 28 in-service teachers who collaborated with GenAI tools to codevelop learning tasks across three weeks. Qualitative analyses identified four overarching interaction challenges: inefficient interactions, low-quality tasks, lack of distributed cognition, and negative human-GenAI feedback loops. These overarching challenges emerged from bilateral deficiencies-three teacher-related shortcomings (providing too little context information, providing too little pedagogical guidance, and lacking GenAI skills) and six GenAI-related shortcomings (accuracy issues, limited prompt understanding, weak knowledge of content sequences, limited context knowledge, incomplete instructional solutions, and usability issues). Quantitative analyses of changes from pre to post showed significant improvements in teachers' technical knowledge and artificial intelligence (AI) literacy, with no changes in self-efficacy and pedagogical knowledge. Finally, teachers expressed professional development needs primarily in two areas: AI foundations and applications and AI pedagogy. This study provides empirical evidence of bidirectional challenges in teacher-GenAI collaboration in designing learning tasks and offers a framework for understanding teacher professional growth through AI collaboration and further needs.

Active Few-Shot Learning (AFSL) adapts LLMs to specialized domains by identifying the most valuable unlabeled samples for annotation and use as few-shot demonstrations, effectively reducing human annotation costs while promoting high performance. However, existing methods typically rely on output-level signals for the sample identification, such as predictive entropy or semantic similarities with test-time data based on external embeddings, which often overlook models’ internal dynamics which could pinpoint specific knowledge gaps. To bridge this gap, we propose NeuFS, a Neuron-Aware Active Few-Shot Learning framework that shifts the selection paradigm from output-level proxies to models’ internal dynamics. NeuFS utilizes neuron activation patterns to represent sample directly, and includes a dual-criteria selection strategy that: (1) ensures few-shot sample diversity with neuron patterns for broader example coverage, while (2) prioritizing on identifying informative and challenging few-shot samples LLMs tend to hallucinate by quantifying neuron consensus. Experiments on three datasets demonstrate that NeuFS excels in both reasoning and text classification tasks, outperforming existing AFSL baselines. Ablation studies further highlight that internal neuron activations provide a more principled and effective selection signal than external embeddings, validating the superiority of the proposed NeuFS.

As quantum information science advances and the need for pre-college engagement grows, a critical question remains: How can young learners be prepared to participate in a field so radically different from what they have encountered before? This paper argues that meeting this challenge will require strong interdisciplinary collaboration with the Learning Sciences (LS), a field dedicated to understanding how people learn and designing theory-guided environments to support learning. Drawing on lessons from previous STEM education efforts, we discuss two key contributions of the learning sciences to quantum information science (QIS) education. The first is design-based research, the signature methodology of learning sciences, which can inform the development, refinement, and scaling of effective QIS learning experiences. The second is a framework for reshaping how learners reason about, learn and participate in QIS practices through shifts in knowledge representations that provide new forms of engagement and associated learning. We call for a two-way partnership between quantum information science and the learning sciences, one that not only supports learning in quantum concepts and practices but also improves our understanding of how to teach and support learning in highly complex domains. We also consider potential questions involved in bridging these disciplinary communities and argue that the theoretical and practical benefits justify the effort.

Generative Artificial Intelligence (GenAI) has sparked a global debate on its potential as a feedback source for students, yet research in this area remains limited. This study explores students’ use of GenAI during peer feedback provision. Fifty-four graduate students enrolled in a master’s course in the food science domain at a Dutch university received instruction on the effective and ethical use of GenAI. They then wrote an argumentative essay, provided feedback to peers, and revised their essays. Finally, students completed an online questionnaire regarding their perceptions and use of GenAI for peer feedback provision. Descriptive analyses were applied to survey data and comment data were coded quantitatively for the presence of comment features. The results revealed that just over half of the students chose not to use GenAI for peer feedback provision, primarily because they believed they would learn more by completing the task independently. The remaining students used GenAI to improve both high-level and low-level aspects of their feedback, and most of these students found GenAI to be moderately helpful for peer feedback provision. In terms of its impact on the peer feedback content, students who used GenAI provided more suggestions for high-level issues and offered less mitigating praise for low-level issues compared to those who did not use GenAI for peer feedback provision. These results offer valuable insights for the design and adoption of GenAI tools to enhance peer feedback practices.

This study breaks new ground by presenting a new, more sophisticated model of learning engagement that goes beyond the current state of the art embodied in the widely used Affective-Behavioral-Cognitive (ABC) model. This work synthesizes and builds upon neglected lines of research in the structure of affective engagement. It also integrates entirely novel theoretical considerations in the form of activity spaces—the different learning spaces associated with different course-defined activities, which in turn afford different forms of cognitive and behavioral engagement. To empirically test these theoretical notions, data from a sample of 655 students across multiple sections of a course were analyzed using structural equation modeling that linked different elements of learning engagement to academic performance outcomes like exam grades, quizzes, participation, and assignment performance. The model fit and practical implications of the traditional ABC structure was compared to that of the Revised Affective-Behavioral-Cognitive model (ABC +) proposed here. The traditional ABC model evinced substantially more misfit to the data (CFI = 0.97, TLI = 0.94, RMSEA = 0.037, [90

Teachers are increasingly called to enact student-centered instructional approaches that promote a socially and intellectually ambitious vision of classroom teaching. Such approaches require adaptive expertise to effectively elicit and respond to students' developing thinking amidst a constantly evolving and multifaceted classroom environment. However, the field's understanding of the cognitive learning mechanisms that support developing adaptive teaching expertise and how to cultivate them in professional learning contexts is still underdeveloped. In this article, we integrate key cognitive and situated learning perspectives, with a particular focus on characterizing in detail the kinds of learning interactions that support teachers' adaptive problem solving and knowledge transfer. Our primary contribution is to offer a revised theory of frames and framing for developing adaptive teaching expertise, situated in the context of expert-guided teacher reflection. Specifically, we describe key components of a teaching frame (teaching schema and teaching principle) and argue that two framing dimensions (strategic and expansive) are especially consequential for facilitating learning interactions that support developing adaptive teaching expertise. Our core argument is that frames and framing are central for understanding teacher learning and cognition, and that extending this literature into teacher learning research holds promise for advancing more robust teacher professional development.

Peer feedback has proven to be practically helpful for students in achieving learning outcomes, especially through peer feedback among students in matching achievement groups. However, some researchers have raised concerns about whether students would be better served by being matched with different achievement groups rather than ‘random’ or same-level matching methods. No studies have investigated whether other ways of matching groups would change outcomes. The study examined the impact of reviewers' expertise and document quality changes on learning outcomes in the matching group through a large dataset involving three large biology courses. Results reveal that matching groups with high reviewer competence are associated with higher learning outcomes for students with low performance. Although receiving more feedback is often negatively associated with learning, receiving more feedback from those with high reviewer expertise is positively related to learning. The study further suggests that educators can focus on the overall competence of members in the assigned reviewers instead of finding matching individuals. In addition, educators and researchers should consider the roles of both providing and receiving feedback based on the ability of matching groups. Future research could use questionnaires or experimental methods to examine student motivation and further determine the causal relationship between the relative benefits/risks of different matching groups on learning outcomes.

Sustainable peer feedback practices heavily depend on students' perceptions of the peer feedback task as well as their self-concepts as peer feedback participants. While empirical studies often focus on one of these two aspects, they frequently overlook the relationship between the two. This mixed-method, longitudinal study surveyed 100 first-year undergraduates twice to examine the relationship between their evolving task perceptions and self-concepts during iterative peer feedback. It also analyzed students' initial writing proficiency and prior peer feedback experiences to deepen our understanding of these relationships. Exploratory Factor Analyses revealed three task perception dimensions and three self-concept dimensions related to peer feedback. Further regression analyses identified multiple bidirectional relationships in predicting changes in both. Most notably, doubts about peer feedback strongly predicted changes in self-concept, and self-concept as peer feedback dialoguers predicted changes in attitudes towards peer feedback tasks. Low writing proficiency learners showed more significant improvements in their task perceptions, and learners without peer feedback experience associated high task perceptions with a stronger positive evaluation of themselves as feedback providers. Interviews with students provided converging evidence and additional depth to these quantitative findings.

Abstract Background The demand for engineers in the workforce continues to rise, which requires increased retention and degree completion at the undergraduate level. Engineering educators need to better understand opportunities to retain students in engineering majors. A strong sense of belonging in engineering represents one important contributor to persistence. However, research has not investigated how academic help-seeking behaviors relate to belonging and downstream outcomes, such as persistence in engineering. Interventions to support and develop belonging show promise in increasing student retention, with particularly positive influences on women, Black, Latino/a/x, and indigenous students. As part of a larger research project, a quasi-experimental intervention to develop a classroom ecology of belonging was conducted at a large Midwestern university in a required first-year, second-semester engineering programming course. The 45-min intervention presented students with stories from past students and peers to normalize academic challenges within the ecology of the classroom as typical and surmountable with perseverance, time, and effort. Results With treatment (n = 737) and control (n = 689) participant responses, we investigated how the intervention condition affected students' comfort with seeking academic help and feeling safe being wrong in class as influences on belonging. Using path analysis, a form of structural equation modeling, we measured the influence of these attitudinal variables on belonging and the influence of belonging beyond a student’s grade point average on enrollment as an engineering major the following fall. The path analysis supports the importance of academic help-seeking and feeling safe to be wrong for belonging, as well as the importance of belonging on continued enrollment. A group path analysis compared the treatment and control groups and demonstrated the positive impact of the intervention on enrollment for the treatment participants. Conclusions The analyses demonstrate the importance of academic help-seeking in students’ sense of belonging in the classroom with implications for identifying effective tools to improve students’ sense of belonging through supporting help-seeking behaviors.

While classrooms in the United States serve an increasingly multilingual student population, the de facto medium-of-instruction policy in most schools is English immersion. Additionally, although evidence suggests that student discussion is crucial to conceptual understanding in mathematics learning, multilingual students and classified English learners in particular tend to rarely engage in meaningful, academically-oriented discussion throughout their school day. In order to promote discussions wherein multilingual peers engage in collaborative mathematical meaning-making, it is critical to understand exactly how home language practices may function as students leverage multilingual learning opportunities in English-medium classrooms. This study drew on quantitative and qualitative data collected in the classrooms of 23 middle school mathematics teachers in a predominantly Spanish-speaking community to investigate how students productively used their home language in mathematics class-including the use of Spanish to deeply process mathematical concepts. Findings suggest that students used Spanish in mathematics learning for two distinct purposes-input-output and processing-and that these separate practices worked together to promote deep engagement with mathematical content and concepts. Results further indicate the value of home language use for processing mathematical meanings in particular and point to the importance of multilingual instructional practices in multilingual classrooms.

Networked Improvement Communities, or NICs, are a type of research-practice partnership that have become a popular approach to large-scale educational change. Despite the growing prevalence of NICs in education, however, there is little research on their effects on student outcomes, particularly with attention to their multilevel complexity and whether they lead to more equitable outcomes. In this study, we test an approach for evaluating the effects of network participation on the growth of student reading outcomes in a large NIC relative to a set of matched non-network schools. Overall, schools participating in the NIC showed growth in reading outcomes that outpaced matched schools in the state by an additional 5 %. Benefits varied by school, which could be explained by levels of teacher participation and depth of implementation. In addition, historically disadvantaged students especially benefited from NIC participation.

Peer learning is a promising instructional strategy, particularly in higher education, where increasing class sizes limits teachers’ abilities to effectively support students’ learning. However, its use in a traditional way is not always highly effective, due to, for example, students’ lack of familiarity with strategies such as peer feedback. Recent advancements in educational technologies, including learning analytics and artificial intelligence (AI), offer new pathways to support and enhance peer learning. This editorial introduces a special issue that examines how emerging educational technologies, specifically learning analytics, AI, and multimodal tools, can be thoughtfully integrated into peer learning to improve its effectiveness and outcomes. The six studies featured in this issue present key innovations, including the successful application of AI-supported peer assessment systems, multimodal learning analytics for analyzing collaborative gestures and discourse, gamified online platforms, social comparison feedback tools and dashboards, group awareness tools for collaborative learning, and behavioral indicators of peer feedback literacy. Collectively, these studies show how these technologies can scaffold peer learning processes, enrich the quality and uptake of peer feedback, foster engagement through gamification, promote reflective and collaborative learning, and address peer feedback literacy. However, the issue also identifies underexplored gaps, such as the short-term nature of many interventions, insufficient focus on the role of teachers, limited cultural and equity considerations, and a need for deeper theoretical integration. This editorial argues for a more pedagogically grounded, inclusive, and context-sensitive approach to technology-enhanced peer learning—one that foregrounds student agency, long-term impact, and interdisciplinary collaboration. The contributions of this special issue provide insights to guide future research, design, and practice in advancing peer learning through educational technologies.

Learner engagement is critical to student success in science, technology, engineering, and mathematics (STEM) coursework. However, the character of engagement is still debated. University courses, particularly in STEM, are complex, multi-context environments in which students must cognitively and behaviorally engage in class, recitations/labs, independent study, and high-stakes exams. We hypothesize that students show consistency in how they engage within each of the particular but show much less consistency across these varied contexts. Knowledge of the fine-grained structure of learning engagement can be used to refine pedagogy and improve learning outcomes. Applying new engagement survey instruments that were iteratively developed to be contextually meaningful, we first present an exploratory factor analysis applied to 1,176 students from two different courses and institutions. Then we present a confirmatory factor analysis applied to 772 students in a third course. The findings support a model in which places of engagement (e.g., lecture, studying) are central to engagement’s structure, and further, they suggest that each place of engagement has a dominant characteristic (e.g., predominantly cognitive or behavioral). This discovery prompts a reevaluation of prevailing STEM learning engagement theories and improves our understanding of engagement dynamics -factors crucial for improving undergraduate academic success in STEM.

Producing large volumes of high-quality, timely feedback poses significant challenges to instructors. To address this issue, automation technologies-particularly Large Language Models (LLMs)-show great potential. However, current LLM-based research still shows room for improvement in terms of feedback quality. Our study proposed a multi-agent approach performing "generation, evaluation, and regeneration" (G-E-RG) to further enhance feedback quality. In the first-generation phase, six methods were adopted, combining three feedback theoretical frameworks and two prompt methods: zero-shot and retrieval-augmented generation with chain-of-thought (RAG_CoT). The results indicated that, compared to first-round feedback, G-E-RG significantly improved final feedback across six methods for most dimensions. Specifically:(1) Evaluation accuracy for six methods increased by 3.36% to 12.98% (p < 0.001); (2) The proportion of feedback containing four effective components rose from an average of 27.72% to an average of 98.49% among six methods, sub-dimensions of providing critiques, highlighting strengths, encouraging agency, and cultivating dialogue also showed great enhancement (p < 0.001); (3) There was a significant improvement in most of the feature values (p < 0.001), although some sub-dimensions (e.g., strengthening the teacher-student relationship) still require further enhancement; (4) The simplicity of feedback was effectively enhanced (p < 0.001) for three methods.

There is an ongoing literacy crisis within the US. One practice that supports reading comprehension growth is the use of student-centered routines. However, teachers often face challenges in integrating them into their instructional contexts. Networked improvement communities (NICs), combining organizational routines of improvement science with the coordinated effort of networks, could help teachers meet this integration challenge. Teachers have found value in NICs, but district leaders are seeking evidence of their impact. To help teachers advocate for this approach, this study analyzes 2 years of student outcome and teacher implementation data in a 13-school US-based NIC that focused teachers on a set of student-centered routines. Students with a higher implementing teacher showed benefits comparable to a typical 2.5 months of learning at the national scale. This study also uses an equity-focused lens to study the impact of instructional change efforts on specific student groups without requiring a formal experimental study.

Peer feedback can be highly effective for learning, but only when students give detailed and helpful feedback. Peer feedback systems often support student reviewers through instructor-generated comment prompts that include various scaffolding features. However, there is little research in the context of higher education on which features tend to be used in practice nor to which extent typical uses impact comment length and comment helpfulness. This study explored the relative frequencies of twelve specific features (divided into metacognitive, motivational, strategic, and conceptual scaffolds) that could be included as scaffolding comment prompts and their relationship to comment length and helpfulness. A large dataset from one online peer review system was used, which involved naturalistic course data from 281 courses at 61 institutions. The degree of presence of each feature was coded in the N = 2883 comment prompts in these courses. Since a given comment prompt often contained multiple features, statistical models were used to tease apart the unique relationship of each comment prompt feature with comment length and helpfulness. The metacognitive scaffolds of prompts for elaboration and setting expectations, and the motivational scaffolds of binary questions were positively associated with mean comment length. The strategic scaffolds of requests for strength identification and example were positively associated with mean comment helpfulness. Only the conceptual scaffold of subdimension descriptions were positively associated with both. Interestingly, instructors rarely included the most useful features in comment prompts. The effects of comment prompt features were larger for comment length than comment helpfulness. Practical implications for designing more effective comment prompts are discussed.

Generative AI (GenAI) has gained attention as a new feedback source in education because it can generate human-like text. However, its use in feedback lacks a strong pedagogical framework, which is necessary for effective implementation. This paper addresses this gap. It outlines human-centered feedback challenges, and then explores human and artificial cognition differences, highlighting the need for hybrid intelligence. Next, it positions GenAI feedback within feedback theory and proposes a definition for GenAI feedback. The paper conceptualizes the role of GenAI feedback as either an independent source or as part of a collaborative process with humans referred to as "Hybrid Intelligent Feedback". Building on this conceptualization, it discusses the approaches and principles of hybrid intelligent feedback and then proposes a pedagogical framework that outlines the implementation steps for hybrid intelligent feedback. The paper concludes by describing the pedagogical framework and outlining recommendations for future research on hybrid intelligent feedback.