Learning to Learn via Artificial Intelligence and Learning Analytics

Abstract

This paper addresses the issue of how artificial intelligence and learning analytics are used to help learners improve their self-regulated learning skills, in other words, “learn how to learn”. Two broad approaches are described. The first approach involves artificial intelligence systems that both teach content as well as support self-regulated learning by means of dynamic adjustment to various aspects of the interaction, including dynamic advice and feedback. Such systems are characterised as either purely reactive, explicit or pro-active. The second approach involves learning analytic systems based on artificial intelligence techniques that capture and analyse learning data from various kinds of educational interactions and later feed that back to learners in a form that enables them to visualise and reflect on their learning behaviour. Whilst the first approach is described by reference to the literature, the second approach showcases experimental work conducted by two of the authors. This involved the analysis of audio transcripts of students’ weekly group discussions and the development of different machine learning techniques to capture the challenge moments, triggers of regulation of learning, that would form the basis for feedback implemented in real-world contexts. This informs the future design of the learning analytics and artificial intelligence system to promote awareness and support learners’ subsequent regulation of learning.

Keywords: Self-regulated learninglearning to learnartificial intelligencelearning analytics

Introduction

One of the most important outcomes for an educational institution is that its learners get better at “doing” learning. Central to “learning to learn” is the skill of self-regulated learning where the learner focuses on the processes of their learning in addition to the content of that learning. These processes include planning, goal setting, monitoring and reflection, not just in terms of metacognition, but also in terms of their meta-affective and meta-motivational facets. One of the contributions of artificial intelligence (AI) has been to assist human teachers in developing their learners’ metacognitive, meta-affective and meta-motivational processes. AI has played different roles, each implying a different division of labour between the artificial intelligence (AI) systems and human teachers. These roles lie along a continuum from AI doing both the teaching of content and fostering self-regulated learning to AI solely undertaking the learning analytics: The AI teaching system acts as both a content teacher as well as promoting self-regulated learning. This involves the human teacher introducing the AI system into their teaching and then priming and motivating the learners into using it effectively. The AI system then (i) adaptively chooses learning tasks, (ii) gathers and analyses learner data as the learners work on those tasks, (iii) adaptively reacts to how well the learners progress on the tasks, and crucially (iv) adaptively reacts to the way the learner has gone about their learning. Finally, handing back to the human teacher a summary of how things have gone for the teacher to provoke whatever analysis, discussion and remedial teaching is then needed. The AI Learning Analytics system does not interact directly with the students while they learn but simply gathers and analyses data on the learners’ learning and self-regulated learning, which it then hands over later to the human teacher and/or the learners for analysis and discussion. This data might have been derived from learners working in a classroom, either with or without any computer-based educational technology. This paper is divided into 6 sections. The next section gives a brief overview of the notion of learning to learn, tracing it from early work by Bandura and Zimmerman on self-regulated learning through the development of static and dynamic models to the present day. Section 3 explores one end of the continuum mentioned above, namely the role that AI has played in providing platforms that support students in both their learning and their learning to learn. Sections 4 and 5 concentrate on the other end of the continuum mentioned above, namely AI systems that collect data on students' learning (not necessarily learning with an AI platform) and make use of learning analytic techniques to process that data into forms amenable to interpretation by learners and/or teachers in support of learning to learn. Section 6 offers a conclusion.

Learning How to Learn

In the context of student learning in educational contexts, “learning how to learn” can be regarded as the student improving some aspect of their self-regulated learning skills. Broadly, these skills involve the student thinking about and acting on both the “how” of learning as well as on the “what” to be learnt. For a recent overview of the field, see Tinajero et al. (2024) and for an overview of the role of AI in supporting metacognition (a central component of self-regulated learning), see Azevedo and Aleven (2013).

The AI System Acts as Both a Content Teacher As Well As Promoting Self-

Regulated Learning

Artificial intelligence-based systems in education have a long history of personalising learning by dynamically adjusting (i) the difficulty of the learning task, (ii) the interface features of the learning environment, and (iii) the feedback and help in response to the systems’ estimation of the learner’s degree of understanding and skill, as well as their affective and motivational states. We distinguish three kinds of systems with respect to the degree that the dynamic adjustments address issues of self-regulation. Purely reactive systems dynamically make the kinds of adjustment mentioned above and these may well help the learner progress better, but they do not in themselves directly address the learner’s self- regulation. This is because such adjustments are made without explaining why they are being made and thus the focus for the learner is likely to remain on the task rather than on how they are learning. For example, at an affective level some systems display a screen-based, pedagogical agent playing the role of an empathetic teacher whose facial expression changes from smiling when the learner is working correctly to frowning when the learner makes a mistake (e.g. Dias et al., 2006). This change of facial expression provides an ambiguous signal that something is not quite right but does not indicate what that is. Explicit systems dynamically make adjustments, as above, but also offer some kind of metacognitive, meta-affective or meta-motivational commentary or feedback to the learner. They may also request information from the learner about, for example, their confidence in their answers or their affective state (e.g. Arroyo et al., 2009). Such systems may or may not generate a log file for the human teacher to use in any post-task interaction with the learners. Arroyo et al. (2014) describe the evaluation of a system for teaching mathematics that interacts with learners at the cognitive, affective, metacognitive and motivational levels. For instance, it measures learner effort as well as success and failure. Depending on their progress, for some students it explicitly praises their effort and for others it de-emphasises the importance of success. Based on Dweck’s (2002) theory of learning and mindset, it also explicitly provides feedback to try to persuade students that mathematics is learnable and not impossible. A commonly used method of providing metacognitive feedback is for a system to display a “skillometer” of some kind that indicates the system’s estimation of the learner's progress in acquiring the skills or concepts to be learned (for an early example, see e.g. Koedinger et al., 1997). A contemporary and more elaborated example is Area9 Rhapsode (White & du Boulay, 2024). This system asks learners to make judgements of their learning (JOLs) on each segment of the course being taken and the system relates those JOLs to the learner’s success or otherwise on that segment. The system then generates a diagram contrasting (i) the degree to which the learner was conscious of how much they had understood and of how much they had not yet understood, against (ii) the degree to which the learner had seemed unconscious of how much they had understood and of how much they had not yet understood. These metacognitive diagrams were aimed both at the learners and at the human teacher who would then use them for classroom discussions about the nature of learners’ individual differences in the trajectories of their learning and understanding. Betty’s Brain, is a system designed to help learners develop their understanding of scientific concepts and processes, e.g. the ecology of a river system (Biswas et al., 2016). The learner is provided with various online materials to read and then starts to construct a node and arrow concept map, where arrows represent processes and nodes represent entities, e.g. plants produce dissolved oxygen. The learners can work at their own pace and in any order. They can submit a partial concept map at any time to have it evaluated. If they submit the same concept map for evaluation again, having made no changes to it, they are offered the metacognitive advice to return to the online materials and update their concept map before requesting it to be evaluated again. MetaTutor is one of the most sophisticated systems that addresses the learner and reacts both at the cognitive and at the metacognitive levels (Azevedo et al., 2022). The basic system aims to teach blood circulation, but it also contains 4 agents monitoring the way the learner is managing their learning: (i) “helps students to navigate through the system and orient the students about the task”, (ii) “guides students in setting appropriate sub-goals”, (iii) “helps students to monitor their progress”, and (iv) “helps students deploy SRL learning strategies, such as summarizing and note-taking, making inferences, re-reading, and generating hypotheses”. “For example, students are prompted to self-assess their understanding and are then given a brief quiz. Quiz results allow the PA [agent] to provide feedback according to the calibration between students’ confidence of comprehension and their actual quiz performance. Learners can also self-initiate and express these same system-initiated metacognitive judgments and learning strategies through an SRL palette of actions.” (Azevedo et al., 2022, Page 6) With the advent of generative AI, in particular LLMs such as ChatGPT, there is renewed interest in the roles that they might play in education in general (see, for example, Giannakos et al., 2024; Pulk & Koris, 2025; Wang et al., 2025) and in support of self-regulated learning in particular. For example, Lai (2024) argues that “chatbots present the ability to be a virtual tutor in the absence of an instructor, adult, or peer. GenAI chatbots can operate as virtual tutors that engage in personalized conversations with learners by providing an interactive and adaptive learning environment to answer questions, explain, and offer instant support, assistance, and guidance.” In the terminology of this section, these roles are largely reactive with respect to self-regulation. By contrast, when an instructor was present, Xu et al. (2025) explored how students were assisted to operate when using generative AI tools in an educational technology course. The ChatGPT-based system generated prompts for planning, monitoring and reflection as well as for evaluation, but was not as sophisticated as MetaTutor in terms of personalising the metacognitive prompting to the learning trajectory of the learner. Pro-active systems go beyond Explicit systems but are a work in progress. Rather than simply responding to a learner’s self-regulatory development needs as they manifest in a given learning session, pro-active systems initiate new learning situations designed specifically to develop aspects of the learner’s self-regulated learning skill-set that need attention. For example, in terms of metacognition this might mean exposing a learner to much more open-ended problems than they are used to, in order to strengthen their ability to set goals, monitor progress and reflect in this new context. In terms of meta-affect and meta- motivation, Goldin (2000) argues for: “a curriculum in which affect is considered may include generating problems from the curiosity of students, in order to develop their sense that intense feelings are appropriate, and teaching them to apply heuristics when these feelings occur. Anxiety, fear, and despair (but not puzzlement, bewilderment, and frustration) maybe regarded as essentially undesirable affective states; yet we need to provide appropriate, domain-specific ways for students to handle the negative affect when it (inevitably) occurs. We also need to provide productive experiences with and uses of desirable affect, including the affect of frustration. Students should occasionally experience frustration- not too much, of course-and then be brought back from it, guided to make progress in the problem with processes suggested by the frustration, and feel subsequent pleasure, elation, and satisfaction that is heightened by the earlier experience of frustration. We give students too little experience with intensely positive affective states, rarely connecting them with their own, genuine mathematical achievements.” (Goldin, 2000, pp. 217-218) Such a curriculum might, one day, be implemented in a pro-active AI-based system aiming to improve learners’ meta-affect and meta-motivation. The following sections explore the other end of the continuum mentioned at the start, namely the use of learning analytics to support improvements in self-regulated learning.

How AI and Learning Analytics Can Be Leveraged to Infer Students’

Regulation of Learning, Potentially Supporting Reflective and Effective

Collaborative Learning

Learning analytics (LA) is an emerging field at the intersection of artificial intelligence and education, which aims to measure, collect, analyse and report data “ about learners and their contexts, for purposes of understanding and optimising learning and the environments in which it occurs” (Romero & Ventura, 2020, p.2). This field leverages both learning sciences theories to guide data collection and interpretation while also leveraging data analytics techniques to analyse data, model educationally meaningful constructs and offer insights for educational stakeholders, including students and teachers. This interdisciplinary approach combines the theoretical grounding of learning sciences with the data-driven analytical power of data sciences, offering complementary strengths for understanding and supporting learning processes. Multiple data sources and analytics approaches have been used in the learning analytics field to convert low-level data collected from learning platforms or environments to high-level educationally meaningful constructs. This process is known as a mapping “From clicks to constructs,” offering meaningful insights, aligning with educators’ pedagogical interests. For example, Cukurova et al. (2020) utilised supervised machine learning models, as decision trees, to detect students’ listening, watching, making and speaking behaviours as well as predict groups’ collaborative problem-solving competence from video data. Kent and Cukurova (2020) leveraged social network analysis to analyse the collaborative learning process of learners’ communities from their online discussion interactions. Advances in natural language processing (NLP) help computationally recognise patterns in texts (Dowell & Kovanovic, 2022), for instance, group discourse, which is pervasively available in collaborative settings. Various NLP techniques have been used in extracting features from discourse. The conventional approach involves the use of bags-of-words (Aggarwal & Zhai, 2012), analysing word frequencies, or linguistic features, grammatical structures in sentences, such as part-of-speech (POS). For instance, Emara et al. (2021) leveraged POS trigrams obtained from group discourse to identify group patterns of regulation in collaborative open-ended problem-solving contexts. These NLP-based features can be utilised, particularly in conjunction with machine learning (ML) techniques, to build predictive models from annotated datasets. To illustrate, Zheng et al. (2019) created a semi-auto ML pipeline combining human coding and NLP features to identify self-regulated learning (SRL) vs socially shared regulation of learning (SSRL) constructs from group chats. Large Language Models (LLMs), which are advanced NLP models pretrained on extensive datasets, offer capabilities to understand patterns in human language, enabling them to receive natural language instructions and respond accordingly (Liu et al., 2021), and potentially effectively even when confronted with unseen data (zero-shot learning) or when provided with minimal training (one-shot or few-shot learning). Despite the potential of NLP and recent developments in analytic techniques, there are insufficient studies and pragmatic implementations of NLP in SSRL research (Villa-Torrano et al., 2025). While some studies applied NLP to unravel the regulatory process (Emara et al., 2021; Zheng et al., 2019), none of them target the emergence of the regulatory process, i.e., predicting challenges and identifying challenge dimensions. To understand shared regulation of learning in collaborative settings, data about learners’ situations and regulation could be gathered from students’ interactions to infer targeted constructs. Since collaboration requires learners to engage in regulation of learning (ROL) processes, it involves collective planning, monitoring, evaluating, and taking control of their own learning to achieve set goals (Hadwin & Oshige, 2011). As the group navigates through the collaborative task, members can individually regulate themselves (self-regulated learning, SRL), co-regulate others (co-regulation, Co-RL), and/or collectively regulate as a group (socially shared regulation, SSRL) to overcome the wide range of challenges that may arise (Hadwin et al., 2011; Kreijns et al., 2003). For regulation, group members must first acknowledge the challenge moments that might hinder effective collaboration and develop suitable strategies to overcome these challenges together (Järvenoja & Järvelä 2009; Malmberg et al., 2015). The use of ROL strategies has been linked to higher academic outcomes (Broadbent & Poon, 2015; Richardson, Abraham and Bond, 2012), with research showing groups highly engaged in SSRL processes negotiated shared task perceptions, goals, plans, and strategies (Hadwin et al., 2011; Malmberg et al., 2015) while maintaining positive socio- emotional interactions, or establishing “mutual trust” to overcome challenges collectively (Fransen, et al., 2011). Without awareness of one’s own SRL and where the group stands as a whole, various perceptions of the challenges can lead to inappropriate or misaligned regulatory processes by group members (Gutwin & Greenberg, 2002; Järvelä et al., 2015). However, due to differences in personal socio-historical experiences situated in various contexts (Hadwin et al., 2011), recognising challenge moments and aligning task perceptions and goals to coordinate SSRL strategies may require external regulatory support (Järvelä et al., 2015). Research has shown that groups who failed to recognise challenge moments accurately tend to activate procedural and behavioural (i.e., “routine-level”) strategies, while highly regulated learners focus more on deep-level processes such as cognitive and metacognitive strategies to overcome challenges collectively (Järvelä et al., 2013; Malmberg et al., 2015). Although researchers have begun to develop computer-based pedagogical tools or pedagogical agents to support SRL (Azevedo & Hadwin, 2005; Perry & Winne, 2006), gathering SRL data through macro-level self-reports may not accurately reflect SRL strategies (Rovers et al., 2019). Such global measures also limit the study and support of regulation processes “on the fly” (Winne et al., 2002), hindering real-time and ongoing support for regulation in CL. To mitigate the validity concern and the lack of fine-grained information in self-reports hindering timely regulatory intervention, we propose a first step, contributing to the lack of research in the emergence of SSRL by leveraging NLP models to identify the dimensions of challenge moments exhibited through group discourse. An accurate identification of challenge moments is a crucial prerequisite to the deployment of appropriate ROL strategies, which would support the alignment of strategies amongst group members for successful collaboration (Järvelä et al., 2015). Three different NLP modelling approaches, namely a rule-based, supervised ML, and LLM approach, were compared in terms of their performance.

An AI System for Real-World Collaboration and Regulation of Higher

Education Students: Lessons Learned

Collaboration analytics (CA), a subfield of LA, focuses on collecting, analysing and supporting group interactions by enabling data-driven decision-making for stakeholders (Schneider et al., 2021). Similarly, CA employs multimodal data sources and diverse techniques to infer collaborative constructs and realise group dynamics (Martinez-Maldonado et al., 2021). Based on our prior models built to detect challenge moments and extend them to cover a detection of regulatory processes, these inferred constructs could be presented as feedback by mirroring information back to learners, closing the feedback loop (Martinez-Maldonado et al., 2021), triggering awareness enhancement of groups by fostering mutual understanding and self-reflection (Chen et al., 2024). Empirical studies have shown that mirroring group constructs such as member knowledge, amount of contribution, member interactions and peer perceptions could improve collaboration, including group awareness, discussion, positive climate, satisfaction, participation and performance (Janssen & Bodemer, 2013). Beyond the mirroring of simple information, tools could be extended to initiate metacognitive monitoring by representing current group states in comparison to desirable states and providing suggestive feedback towards productive group interactions (Soller et al., 2005). Due to its flexibility in presenting information, CA holds promise for simultaneously supporting group awareness through a reflection on current group interactions, enhancing metacognitive monitoring of situational challenges and facilitating regulation through suggested strategies—thereby acting as mediating feedback that fosters a transactional loop in collaborative learning. Despite extensive research in understanding collaborative processes, few studies (the exception includes Echeverria et al., 2019) have implemented CA feedback as support and evaluated it with stakeholders, particularly in real-world collocated small-group collaborative settings. Moreover, CA implementation in an ecologically valid setting has several challenges, ranging from technological, design, deployment, and ethical issues (Martinez-Maldonado et al., 2024). This underscores the urgent need for further exploration of the ethical implications of CA, particularly in authentic face-to-face (f2f) settings. To address these gaps, our work proposed the implementation of a collocated CA feedback intervention to externalise group regulations, accompanied by actionable suggestions to ultimately promote group awareness and trigger effective collaboration (as documented in Suraworachet et al., 2025). Through a human-centred design approach, the study investigates a post-hoc evaluation of students’ perceptions of the proposed CA in authentic small-group collaborative environments, exploring the extent to which challenge moments and regulatory processes were externalised, recognised, and triggered regulation among members, leading to an improvement in collaboration. Ethical considerations on the practical application of the CA feedback, along with future suggestions, were also examined to inform the development and implementation of a more ethical and practical version of CA.

Methodology and feedback design

The study was conducted from a post-graduate module in educational technology where students participated in a 1-hour in-person group activity weekly, discussing their educational technological design. They were purposefully assigned to a fixed small group according to their variation in background of study, years of working experience, gender, and English as a first language. The sessions were video and audio recorded for analysing their group interactions and regulation. Due to technical challenges in capturing and analysing data in real-time, this study demonstrates the feasibility of the proposed CA by collecting real-world data from one week to be processed and displayed for feedback in the subsequent week. Following the protocol described in Suraworachet et al. (2024), students’ discourse was processed to predict the dimensions of challenge moments and extend to cover the prediction of regulatory processes. Due to the similar performances of the two language models (SML and LLM) with different explainability features, we applied both techniques in analysing the current dataset for generating feedback. The RF model and GPT4 were selected to represent SML and LLMs, respectively. RF’s feature importance and GPT4’s decision explanations were utilised to investigate students’ perceptions of model explainability. A web application was created to provide the predicted results in the form of a dashboard featuring both visual representations and textual feedback, targeting awareness enhancement in their challenge moments and socially shared regulation of learning. The feedback was designed according to Hattie and Timperley's (2007) and Soller et al.'s (2005) feedback guidance. Feedback on the detected constructs and suggestions included explanations on what students have done well, as detected from data (learning progress) and what needs to be further improved for better collaboration (prospective activities towards learning goals). Collaboration and regulation theories, combined with expert rules, were used to formulate multiple criteria (e.g., detected challenges, etc.) for textual feedback to help students determine directions for improvement. Figure 1 shows the analytics feedback on group challenges and regulations proposed in the study delivered at the students’ discretion. The group regulation feedback started with an introduction of constructs (challenging dimensions and regulatory processes) due to unfamiliar concepts of regulation. Bar charts were utilised to present the frequency of self-reported preparation and motivation levels vs detected challenges and regulations in the session, whereas a timeline was used to present detected challenges and regulations over time. The textual feedback criteria include (1) a prevalence of specific dimensions of challenges in conjunction with self-report data on the level of preparation or motivation, and (2) regulatory patterns found in learning sessions. For example, if cognitive challenges are prevalent in conjunction with a low level of reported preparation, students receive feedback suggesting higher preparation for the next sessions. Similarly, if there is a low self-reported motivation level alongside high motivational/emotional challenges, the feedback would attribute the cause of high motivational challenges to low motivation among members, suggesting the promotion of a positive group climate. Additionally, to promote student agency, aligning with regulatory skills, model transparency and explainability were explored within this context. The feedback interface allowed students to select a model (either SML or GPT4), which then populated the results accordingly. The actual transcript of a learning session, along with predicted results and decision explanations, was included to promote students’ model inspection and evidence-based reasoning. While GPT4 generated a sentence explaining its reasoning, the SML model outputted a list of important features dominating the model decision, as shown in Figure 1(b).

(a) (b)

See Full Size >

See Full Size >

To investigate students’ perceptions of the feedback, students’ weekly reflections and follow-up voluntary focus groups were utilised and qualitatively analysed.

Conclusion

This paper has contrasted two approaches for using artificial intelligence to develop students’ self- regulated learning capability. In one approach the artificial intelligence system takes on two roles: (i) teaching the content, making dynamic changes to task selection, interface features, provision of help and feedback with respect to the learner’s understanding of that content, and (ii) to monitor how the learner is going about their learning and offer dynamic feedback and advice as to how they might be more effective in their learning. The other approach is to separate the monitoring of and reacting to both the progress in understanding that content and the progress in getting better at learning from the teaching of the content. This involves using artificial intelligence techniques to analyse data from the learning, typically after the event, and transforming it into an actionable visualisation that the learner can use to reflect on how they went about their learning with a view to improving their learning methods the next time. The first approach has to deal with the issue of dividing the learner’s attention between thinking about the content and thinking about their learning. The second approach solves that problem but then faces the issue of the learner needing to relate the post-hoc visualisation of other learning to when their experience when learning occurred, and then carrying forward any insights from their reflection to future learning opportunities.

Acknowledgments

We would like to thank DUTE students for granting permission to collect data for this study. We also extend our gratitude to Snowflake Inc. for providing additional resource space on Streamlit Cloud to support our project.

Conflict of Interest Statement

The authors declare that they have no competing interests.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy or ethical restrictions.

Ethical Statement

Ethical approval for this study was obtained from the IOE Research Ethics Committee (IOE UCL’s Faculty of Education and Society) (Approval Code: Z6364106/2024/06/68 social research). Informed consent from participants was acquired prior to data collection.

Funding

This research is part of the Teacher-AI Complementarity (TAICo) project funded by the European Union’s Horizon Programme under the HORIZON-CL2-2024-TRANSFORMATIONS-01 call with the Project ID: 101177268.

References

• Aggarwal, C. C., & Zhai, C. (2012). An Introduction to Text Mining. In C. C. Aggarwal & C. Zhai (Eds), Mining Text Data (pp. 1–10). Springer US. https://doi.org/10.1007/978-1-4614-3223-4_1

  Google Scholar

  Crossref

• Arroyo, I., Cooper, D. G., Burleson, W., Woolf, B. P., Muldner, K., & Christopherson, R. (2009). Emotion Sensors Go to School. In V. Dimitrova, R. Mizoguchi, B. du Boulay, & A. Graesser (Eds.), Artificial Intelligence in Education. Building Learning Systems that Care: from Knowledge Representation to Affective Modelling (Vol. Frontiers in Artificial Intelligence and Applications 200, pp. 17-24)

• Arroyo, I., Woolf, B. P., Burleson, W., Muldner, K., Rai, D., & Tai, M. (2014). A Multimedia Adaptive Tutoring System for Mathematics that Addresses Cognition, Metacognition and Affect. International Journal of Artificial Intelligence in Education, 24(4), 387-426. https://doi.org/10.1007/s40593-014-0023-y

  Google Scholar

  Crossref

• Azevedo, R., & Aleven, V. (2013). International Handbook of Metacognition and Learning Technologies. Springer-Verlag

• Azevedo, R., & Hadwin, A. F. (2005). Scaffolding Self-regulated Learning and Metacognition – Implications for the Design of Computer-based Scaffolds. Instructional Science, 33. https://doi.org/10.1007/s11251-005-1272-9

  Google Scholar

  Crossref

• Azevedo, R., Bouchet, F., Duffy, M., Harley, J., Taub, M., Trevors, G., Cloude, E., Dever, D., Wiedbusch, M., Wortha, F., & Cerezo, R. (2022). Lessons Learned and Future Directions of MetaTutor: Leveraging Multichannel Data to Scaffold Self-Regulated Learning With an Intelligent Tutoring System. Frontiers in Psychology, 13. https://doi.org/10.3389/fpsyg.2022.813632

  Google Scholar

  Crossref

• Bakhtiar, A., & Hadwin, A. (2020). Dynamic Interplay between Modes of Regulation During Motivationally Challenging Episodes in Collaboration. Frontline Learning Research, 8(2), 1-34. https://doi.org/10.14786/flr.v8i2.561

  Google Scholar

  Crossref

• Bandura, A. (1977). Self-efficacy: Toward a Unifying Theory of Behavioral Change. Psychological review, 84(2), 191-215. https://doi.org/10.1037/0033-295X.84.2.191

  Google Scholar

  Crossref

• Biswas, G., Segedy, J. R., & Bunchongchit, K. (2016). From Design to Implementation to Practice a Learning by Teaching System: Betty’s Brain. International Journal of Artificial Intelligence in Education, 26(1), 350-364. https://doi.org/http://dx.doi.org/10.1007/s40593-015-0057-9

  Google Scholar

  Crossref

• Broadbent, J., & Poon, W. L. (2015). Self-regulated learning strategies & academic achievement in online higher education learning environments: A systematic review. The Internet and Higher Education, 27, 1-13. https://doi.org/10.1016/j.iheduc.2015.04.007

  Google Scholar

  Crossref

• Chen, D., Zhang, Y., Luo, H., Zhu, Z., Ma, J., & Lin, Y. (2024). Effects of group awareness support in CSCL on students’ learning performance: A three-level meta-analysis. International Journal of Computer-Supported Collaborative Learning, 19(1), 97-129. https://doi.org/10.1007/s11412-024-09418-3

  Google Scholar

  Crossref

• Cukurova, M. (2024). The interplay of learning, analytics and artificial intelligence in education: A vision for hybrid intelligence. In British Journal Of Educational Technology . Wiley. https://doi.org/10.1111/bjet.13514

  Google Scholar

  Crossref

• Cukurova, M., Luckin, R., & Kent, C. (2020). Impact of an Artificial Intelligence Research Frame on the Perceived Credibility of Educational Research Evidence. International Journal of Artificial Intelligence in Education, 30(2), 205-235. https://doi.org/10.1007/s40593-019-00188-w

  Google Scholar

  Crossref

• Cukurova, M., Zhou, Q., Spikol, D., & Landolfi, L. (2020). Modelling collaborative problem-solving competence with transparent learning analytics: Is video data enough? Proceedings of the Tenth International Conference on Learning Analytics & Knowledge , 270–275. https://doi.org/10.1145/3375462.3375484

  Google Scholar

  Crossref

• D’Mello, S. K., Duran, N., Michaels, A., & Stewart, A. E. B. (2024). Improving collaborative problem- solving skills via automated feedback and scaffolding: A quasi-experimental study with CPSCoach 2.0. User Modeling and User-Adapted Interaction . https://doi.org/10.1007/s11257-023-09387-6

  Google Scholar

  Crossref

• Dias, J., Paiva, A., Vala, M., Aylett, R., Woods, S., Zoll, C., & Hall, L. (2006). Empathic characters in computer-based personal and social education. In M. Pivec (Ed.), Affective and Emotional Aspects Of Human-Computer Interaction: Game-Based and Innovative Learning Approaches (pp. 246-254)

• Dowell, N., & Kovanovic, V. (2022). Chapter 11: Modeling Educational Discourse with Natural Language Processing. In C. Lang, G. Siemens, A. F. Wise, D. Gasevic, & A. Merceron (Eds), Handbook of Learning Analytics (2nd edn, p. 15). SoLAR. https://doi.org/10.18608/hla22

  Google Scholar

  Crossref

• Dweck, C. S. (2002). Messages that motivate: How praise molds students' beliefs, motivation, and performance (in surprising ways). In J. M. Aronson (Ed.), Improving academic achievement: impact of psychological factors on education (pp. 37-60)

• Echeverria, V., Martinez-Maldonado, R., & Buckingham Shum, S. (2019). Towards Collaboration Translucence: Giving Meaning to Multimodal Group Data. Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems - CHI ’19, 1. https://doi.org/10.1145/3290605.3300269

  Google Scholar

  Crossref

• Efklides, A. (2011). Interactions of Metacognition with Motivation and Affect in Self-Regulated Learning: The MASRL Model. Educational psychologist, 46(1), 6-25. https://doi.org/10.1080/00461520.2011.538645

  Google Scholar

  Crossref

• Efklides, A., & Schwartz, B. L. (2024). Revisiting the Metacognitive and Affective Model of Self- Regulated Learning: Origins, Development, and Future Directions. Educational Psychology Review, 36(2). https://doi.org/10.1007/s10648-024-09896-9

  Google Scholar

  Crossref

• Emara, M., Hutchins, N., Grover, S., Snyder, C., & Biswas, G. (2021). Examining Student Regulation of Collaborative, Computational, Problem-Solving Processes in Open-Ended Learning Environments. Journal of Learning Analytics, 8(1), 49-74. https://doi.org/10.18608/jla.2021.7230

  Google Scholar

  Crossref

• Fransen, J., Kirschner, P. A., & Erkens, G. (2011). Mediating team effectiveness in the context of collaborative learning: The importance of team and task awareness. Group Awareness in CSCL Environments, 27(3), 1103-1113. https://doi.org/10.1016/j.chb.2010.05.017

  Google Scholar

  Crossref

• Giannakos, M., Azevedo, R., Brusilovsky, P., Cukurova, M., Dimitriadis, Y., Hernandez-Leo, D., Järvelä, S., Mavrikis, M., & Rienties, B. (2024). The promise and challenges of generative AI in education. Behaviour & Information Technology, 1. https://doi.org/10.1080/0144929x.2024.2394886

  Google Scholar

  Crossref

• Goldin, G. A. (2000). Affective Pathways and Representation in Mathematical Problem Solving. Mathematical Thinking and Learning, 2(3), 209-219. https://doi.org/10.1207/S15327833MTL0203_3

  Google Scholar

  Crossref

• Gutwin, C., & Greenberg, S. (2002). A Descriptive Framework of Workspace Awareness for Real-Time Groupware. Computer Supported Cooperative Work (CSCW), 11. https://doi.org/10.1023/A:1021271517844

  Google Scholar

  Crossref

• Hadwin, A. F., Bakhtiar, A., & Miller, M. (2018). Challenges in online collaboration: Effects of scripting shared task perceptions. International Journal of Computer-Supported Collaborative Learning, 13(3), 301-329. https://doi.org/10.1007/s11412-018-9279-9

  Google Scholar

  Crossref

• Hadwin, A., & Oshige, M. (2011). Self-Regulation, Co-regulation, and Socially Shared Regulation: Exploring Perspectives of Social in Self-Regulated Learning Theory. Teachers College Record, 113(2), 240-264. https://doi.org/10.1177/016146811111300204

  Google Scholar

  Crossref

• Hadwin, A., Järvelä, S., & Miller, M. (2011). Self-Regulated, Co-Regulated, and Socially Shared Regulation of Learning. In Handbook of Self-Regulation of Learning and Performance (pp. 79–98). Routledge. https://doi.org/10.4324/9780203839010.ch5

  Google Scholar

  Crossref

• Hadwin, A., Järvelä, S., & Miller, M. (2017). Self-regulation, co-regulation, and shared regulation in collaborative learning environments. In Handbook of self-regulation of learning and performance (pp. 83–106)

• Hattie, J., & Timperley, H. (2007). The Power of Feedback. Review of Educational Research, 77(1), 81-112. https://doi.org/10.3102/003465430298487

  Google Scholar

  Crossref

• Howell, J. A., Roberts, L. D., & Mancini, V. O. (2018). Learning analytics messages: Impact of grade, sender, comparative information and message style on student affect and academic resilience. Computers in Human Behavior, 89, 8-15. https://doi.org/10.1016/j.chb.2018.07.021

  Google Scholar

  Crossref

• Janssen, J., & Bodemer, D. (2013). Coordinated Computer-Supported Collaborative Learning: Awareness and Awareness Tools. Educational Psychologist, 48(1), 40-55. https://doi.org/10.1080/00461520.2012.749153

  Google Scholar

  Crossref

• Järvelä, S., & Hadwin, A. F. (2013). New Frontiers: Regulating Learning in CSCL. Educational psychologist, 48(1), 25-39. https://doi.org/10.1080/00461520.2012.748006

  Google Scholar

  Crossref

• Järvelä, S., & Järvenoja, H. (2011). Socially Constructed Self-Regulated Learning and Motivation Regulation in Collaborative Learning Groups. Teachers College Record: The Voice of Scholarship in Education, 113(2), 350-374. https://doi.org/10.1177/016146811111300205

  Google Scholar

  Crossref

• Järvelä, S., Järvenoja, H., Malmberg, J., & Hadwin, A. F. (2013). Exploring socially shared regulation in the context of collaboration. Journal of Cognitive Education and Psychology, 12(3), 267-286. https://doi.org/10.1891/1945-8959.12.3.267

  Google Scholar

  Crossref

• Järvelä, S., Kirschner, P. A., Panadero, E., Malmberg, J., Phielix, C., Jaspers, J., Koivuniemi, M., & Järvenoja, H. (2015). Enhancing socially shared regulation in collaborative learning groups: Designing for CSCL regulation tools. Educational Technology Research and Development, 63, 125-142. https://doi.org/10.1007/s11423-014-9358-1

  Google Scholar

  Crossref

• Järvenoja, H., & Järvelä, S. (2009). Emotion control in collaborative learning situations: Do students regulate emotions evoked by social challenges. British Journal of Educational Psychology, 79(3), 463-481. https://doi.org/10.1348/000709909X402811

  Google Scholar

  Crossref

• Jivet, I., Wong, J., Scheffel, M., Valle Torre, M., Specht, M., & Drachsler, H. (2021). Quantum of Choice: How learners’ feedback monitoring decisions, goals and self-regulated learning skills are related. LAK21: 11th International Learning Analytics and Knowledge Conference, 416. https://doi.org/10.1145/3448139.3448179

  Google Scholar

  Crossref

• Kent, C., & Cukurova, M. (2020). Investigating Collaboration as a Process with Theory-driven Learning Analytics. Journal of Learning Analytics, 7(1), 59-71. https://doi.org/10.18608/jla.2020.71.5

  Google Scholar

  Crossref

• Koedinger, K. R., Anderson, J. R., Hadley, W. H., & Mark, M. A. (1997). Intelligent Tutoring Goes to School in the Big City. International Journal of Artificial Intelligence in Education, 8(1), 30-43

• Kreijns, K., Kirschner, P. A., & Jochems, W. (2003). Identifying the pitfalls for social interaction in computer-supported collaborative learning environments: A review of the research. Computers in Human Behavior, 19(3), 335-353. https://doi.org/10.1016/S0747-5632(02)00057-2

  Google Scholar

  Crossref

• Lai, J. W. (2024). Adapting Self-Regulated Learning in an Age of Generative Artificial Intelligence Chatbots. Future Internet, 16(6). https://doi.org/10.3390/fi16060218

  Google Scholar

  Crossref

• Malmberg, J., Järvelä, S., Järvenoja, H., & Panadero, E. (2015). Promoting socially shared regulation of learning in CSCL: Progress of socially shared regulation among high-and low-performing groups. Computers in Human Behavior, 52, 562-572. https://doi.org/10.1016/j.chb.2015.03.082

  Google Scholar

  Crossref

• Martínez-López, Z., Nouws, S., Villar, E., Mayo, M. E., & Tinajero, C. (2023). Perceived social support and self-regulated learning: A systematic review and meta-analysis. International Journal of Educational Research Open, 5. https://doi.org/10.1016/j.ijedro.2023.100291

  Google Scholar

  Crossref

• Martinez-Maldonado, R., Echeverria, V., Fernandez Nieto, G., & Buckingham Shum, S. (2020). From Data to Insights: A Layered Storytelling Approach for Multimodal Learning Analytics. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1. https://doi.org/10.1145/3313831.3376148

  Google Scholar

  Crossref

• Martinez-Maldonado, R., Echeverria, V., Fernandez-Nieto, G., Yan, L., Zhao, L., Alfredo, R., Li, X., Dix, S., Jaggard, H., Wotherspoon, R., Osborne, A., Shum, S. B., & Gašević, D. (2024). Lessons Learnt from a Multimodal Learning Analytics Deployment In-the-Wild. ACM Transactions on Computer-Human Interaction, 31(1), 1-41. https://doi.org/10.1145/3622784

  Google Scholar

  Crossref

• Miele, D. B., & Scholer, A. A. (2017). The Role of Metamotivational Monitoring in Motivation Regulation. Educational psychologist, 53(1), 1-21. https://doi.org/10.1080/00461520.2017.1371601

  Google Scholar

  Crossref

• Näykki, P., Isohätälä, J., & Järvelä, S. (2021). “You really brought all your feelings out” – Scaffolding students to identify the socio-emotional and socio-cognitive challenges in collaborative learning. Learning, Culture and Social Interaction, 30. https://doi.org/10.1016/j.lcsi.2021.100536

  Google Scholar

  Crossref

• Perry, N. E., & Winne, P. H. (2006). Learning from Learning Kits: gStudy Traces of Students’ Self- Regulated Engagements with Computerized Content. Educ Psychol Rev, 18

• Pintrich, P. R. (2004). A Conceptual Framework for Assessing Motivation and Self-Regulated Learning in College Students. Educational Psychology Review, 16(4). https://doi.org/10.1007/s10648-004-0006-x

  Google Scholar

  Crossref

• Pulk, K., & Koris, R. (2025). Generative AI in Higher Education: The Good, the Bad, and the Ugly

• Raaijmakers, S. F., Baars, M., Paas, F., Merriënboer, J. J., & Van Gog, T. (2018). Training self‐ assessment and task‐selection skills to foster self‐regulated learning: Do trained skills transfer across domains? Applied Cognitive Psychology, 32(2), 270-277. Object Identifier (DOI). https://doi.org/10.1002/acp.3392Digital

  Google Scholar

  Crossref

• Richardson, M., Abraham, C., & Bond, R. (2012). Psychological correlates of university students’ academic performance: A systematic review and meta-analysis. Psychological Bulletin, 138(2), 353-387. https://doi.org/10.1037/a0026838

  Google Scholar

  Crossref

• Romero, C., & Ventura, S. (2020). Educational data mining and learning analytics: An updated survey. WIREs Data Mining and Knowledge Discovery, 10(3). https://doi.org/10.1002/widm.1355

  Google Scholar

  Crossref

• Rovers, S. F. E., Clarebout, G., Savelberg, H. H. C. M., De Bruin, A. B. H., & Van Merriënboer, J. J. G. (2019). Granularity matters: Comparing different ways of measuring self-regulated learning. Metacognition and Learning, 14(1), 1-19. https://doi.org/10.1007/s11409-019-09188-6

  Google Scholar

  Crossref

• Schneider, B., Dowell, N., & Thompson, K. (2021). Collaboration Analytics—Current State and Potential Futures. Journal of Learning Analytics, 8(1), 1-12. https://doi.org/10.18608/jla.2021.7447

  Google Scholar

  Crossref

• Soller, A., Martínez-Monés, A., Jermann, P., & Muehlenbrock, M. (2005). From Mirroring to Guiding: A Review of State of the Art Technology for Supporting Collaborative Learning. I. J. Artificial Intelligence in Education, 15, 261-290

• Stahl, G. (2004). Building Collaborative Knowing. In J.-W. Strijbos, P. A. Kirschner, & R. L. Martens (Eds), What We Know About CSCL: And Implementing It In Higher Education (pp. 53–85). Springer Netherlands. https://doi.org/10.1007/1-4020-7921-4_3

  Google Scholar

  Crossref

• Suraworachet, W., Seon, J., & Cukurova, M. (2024). Predicting challenge moments from students’ discourse: A comparison of GPT-4 to two traditional natural language processing approaches. Proceedings of the 14th Learning Analytics and Knowledge Conference, 473. https://doi.org/10.1145/3636555.3636905

  Google Scholar

  Crossref

• Suraworachet, W., Zhou, Q., & Cukurova, M. (2025). University Students’ Perceptions of a Multimodal AI System for Real‐World Collaboration Analytics: Lessons Learned From a Case Study. Journal of Computer Assisted Learning, 41(5). https://doi.org/10.1111/jcal.70103

  Google Scholar

  Crossref

• Tinajero, C., Mayo, M. E., Villar, E., & Martinez-Lopez, Z. (2024). Classic and modern models of self- regulated learning: integrative and componential analysis. Frontiers in Psychology, 15. https://doi.org/10.3389/fpsyg.2024.1307574

  Google Scholar

  Crossref

• Ucan, S., & Webb, M. (2015). Social Regulation of Learning During Collaborative Inquiry Learning in Science: How does it emerge and what are its functions? International Journal of Science Education , 37 (15), 2503–2532. https://doi.org/10.1080/09500693.2015.1083634

  Google Scholar

  Crossref

• Villa-Torrano, C., Suraworachet, W., Gómez-Sánchez, E., Asensio-Pérez, J. I., Bote-Lorenzo, M. L., Martínez-Monés, A., Zhou, Q., Cukurova, M., & Dimitriadis, Y. (2025). Using learning design and learning analytics to promote, detect and support Socially-Shared Regulation of Learning: A systematic literature review. Computers & Education, 105261. https://doi.org/10.1016/j.compedu.2025.105261

  Google Scholar

  Crossref

• Wang, P., Jing, Y., & Shen, S. (2025). A systematic literature review on the application of generative artificial intelligence (GAI) in teaching within higher education: Instructional contexts, process, and strategies. The Internet and Higher Education, 65. https://doi.org/10.1016/j.iheduc.2025.100996

  Google Scholar

  Crossref

• White, J., & Boulay, B. (2024). Supporting Learners’ Metacogniton and Meta-Affect. In F. Santoianni, G. Giannini, & A. Ciasullo (Eds.), Mind, Body, and Digital Brains (pp. 59-79). Springer, Cham. https://doi.org/10.1007/978-3-031-58363-6_5

  Google Scholar

  Crossref

• Winne, P. H., & Hadwin, A. F. (1998). Studying as self-regulated learning. In D. J. Hacker & J. Dunlosky (Eds.), Metacognition in educational theory and practice

• Winne, P. H., Jamieson-Noel, D., & Muis, K. (2002). Methodological issues and advances in researching tactics, strategies, and self-regulated learning. Advances in Motivation and Achievement: New Directions in Measures and Methods, 12, 121-155

• Wise, A. F., & Schwarz, B. B. (2017). Visions of CSCL: eight provocations for the future of the field. International Journal of Computer-Supported Collaborative Learning, 12(4), 423-467. https://doi.org/10.1007/s11412-017-9267-5

  Google Scholar

  Crossref

• Xu, X., Qiao, L., Cheng, N., Liu, H., & Zhao, W. (2025). Enhancing self‐regulated learning and learning experience in generative AI environments: The critical role of metacognitive support. British Journal of Educational Technology, 56(5), 1842-1863. https://doi.org/10.1111/bjet.13599

  Google Scholar

  Crossref

• Zheng, J., Xing, W., & Zhu, G. (2019). Examining sequential patterns of self- and socially shared regulation of STEM learning in a CSCL environment. Computers & Education, 136, 34-48. https://doi.org/10.1016/j.compedu.2019.03.005

  Google Scholar

  Crossref

• Zimmerman, B. J. (1989). A Social Cognitive View of Self-Regulated Academic Learning. Journal of educational psychology, 81(3), 329-339. https://doi.org/10.1007/978-1-4612-3618-4_1

  Google Scholar

  Crossref