pith. sign in

arxiv: 2605.04816 · v2 · pith:7KBVAEJNnew · submitted 2026-05-06 · 💻 cs.HC

Building AI Companions that Prioritise Learning over Performance

Pith reviewed 2026-05-19 17:34 UTC · model grok-4.3

classification 💻 cs.HC
keywords AI learning companionslearning-performance paradoxpedagogical designadaptive AIeducational technologymetacognitive growthlearner agencyLLM in education
0
0 comments X

The pith

AI in education must move from performance-boosting LLMs to deliberately designed learning companions that prioritize durable understanding.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper identifies a learning-performance paradox in which LLMs improve short-term task outputs like writing or coding yet can weaken cognitive growth, knowledge transfer, and metacognitive skills. It proposes AI learning companions as adaptive, LLM-powered agents built specifically for learning environments rather than general performance. A framework with three foundations guides their design: a pedagogical one that addresses how students learn with AI, an adaptive one that lets the AI learn about the student, and a responsible one that enforces transparency, accountability, inclusivity, and security. Five case studies across different contexts illustrate both the potential and the gaps in current tools. A sympathetic reader cares because this shifts the goal from faster answers to lasting learner agency and growth.

Core claim

The paper claims that LLMs should be redeveloped as AI learning companions rather than prompted as tutors; these companions integrate a pedagogical foundation on student-AI learning interactions, an adaptive foundation on AI modeling of learners, and a responsible design foundation for transparency and security, with the aim of supporting metacognitive growth, knowledge transfer, and learner agency instead of only immediate performance.

What carries the argument

The three-foundation framework for AI learning companions that combines pedagogy, adaptation to the learner, and responsible design principles.

If this is right

  • AI companions can promote metacognitive development and learner agency by design rather than by accident.
  • These systems adapt to individual students instead of offering uniform task assistance.
  • Responsible design keeps educational AI transparent, accountable, inclusive, and secure.
  • Current tools show partial promise but reveal limitations that the framework can address.
  • Integration across varied educational levels and contexts supports durable understanding over quick outputs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Adoption of the framework might require new evaluation metrics that track transfer and agency instead of task speed.
  • Teachers could gain roles as guides who monitor how companions interact with students rather than direct instructors.
  • The approach opens questions about how to balance AI support with opportunities for productive struggle in learning.

Load-bearing premise

That embedding the three design foundations into LLM agents will actually resolve the paradox and produce measurable gains in cognitive growth and knowledge transfer in real settings.

What would settle it

A randomized classroom study that measures long-term knowledge retention and metacognitive skills after using standard LLMs versus companions built on the proposed framework and finds no advantage or a disadvantage for the companions.

Figures

Figures reproduced from arXiv: 2605.04816 by Dragan Gasevic, Hassan Khosravi, Jason Lodge, Jason Tangen, Kristen DiCerbo, Lixiang Yan, Paul Denny, Ryan S. Baker, Shazia Sadiq, Simon Buckingham Shum.

Figure 2
Figure 2. Figure 2: Framework for building AI companions to support learning. The left cycle represents how view at source ↗
Figure 3
Figure 3. Figure 3: Examples of Khanmigo operating in different interaction modes. Examples of Khanmigo operating in different view at source ↗
Figure 4
Figure 4. Figure 4: Illustrative overview of the RiPPLE platform and its AI-assisted learning workflows. Panel (a) presents view at source ↗
Figure 5
Figure 5. Figure 5: Illustrative examples from the CodeHelp platform. Panel (a) illustrates a guardrailed response in the ‘Code view at source ↗
Figure 6
Figure 6. Figure 6: Overview of the JeepyTA workflow 6.4.2 Pedagogical Foundations of JeepyTA Deep and Interactive Learning. JeepyTA engages students through sustained dialogue. When responding to reflections on readings and lectures, it acknowledges contributions, reinforces key ideas, and connects insights to course themes. It clarifies concepts by summarizing arguments and citing specific readings. When students introduce … view at source ↗
read the original abstract

Large language models (LLMs) are rapidly transforming knowledge work by improving the quality and efficiency of tasks such as writing, coding, and data analysis. However, their growing use in education has exposed a learning-performance paradox: while they can enhance short-term task performance, they may also undermine genuine learning, including cognitive growth, knowledge transfer, and metacognitive development. This paper addresses the question of how artificial intelligence should be designed and used to support learning rather than merely improve immediate outputs. We introduce the concept of AI learning companions, defined as adaptive, pedagogically informed, LLM-powered agents designed for integration into learning environments. We propose a framework for their design built on three interrelated foundations: a pedagogical foundation focused on how students learn with AI, an adaptive foundation focused on how AI learns about students, and a responsible design foundation ensuring systems remain transparent, accountable, inclusive, and secure. The framework is illustrated through five case studies spanning diverse educational contexts, levels, and tool designs, revealing both the promise and current limitations of existing tools. We conclude that there is a necessary shift away from LLMs designed for task-oriented performance, and beyond simply prompting them to act as tutors, toward deliberately developed AI learning companions that are pedagogically sound, adapt to their learners, and foster durable understanding, metacognitive growth, and learner agency.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper identifies a learning-performance paradox in LLM use for education, where short-term task performance improves but genuine learning, cognitive growth, knowledge transfer, and metacognition may be undermined. It introduces AI learning companions as adaptive, pedagogically informed LLM-powered agents and proposes a framework built on three foundations: pedagogical (focused on how students learn with AI), adaptive (focused on how AI learns about students), and responsible design (ensuring transparency, accountability, inclusivity, and security). The framework is illustrated through five case studies spanning diverse educational contexts, levels, and tool designs, leading to the conclusion that a necessary shift is required away from task-oriented LLMs toward deliberately developed companions that foster durable understanding, metacognitive growth, and learner agency.

Significance. If the proposed foundations can be operationalized with concrete mechanisms and subjected to empirical testing, the work could have notable significance for HCI and educational technology by providing a structured conceptual lens for designing AI systems that prioritize long-term learning outcomes. It usefully synthesizes pedagogical principles with adaptive AI capabilities and responsible design considerations, potentially informing future tool development and research agendas on AI companions.

major comments (2)
  1. [Case Studies] Case Studies section: The five case studies review existing tools and note limitations but supply no new controlled data, implementation details, or outcome metrics showing that the proposed pedagogical, adaptive, and responsible foundations actually improve knowledge transfer or learner agency over standard LLM use. This leaves the central claim that the framework resolves the learning-performance paradox unexamined and untested.
  2. [Framework] Framework section: The adaptive foundation is outlined at a high level as focusing on how AI learns about students, yet no specific mechanisms (such as student modeling approaches, dynamic feedback loops, or methods for encoding pedagogical principles into LLM behavior) are detailed. This specificity is load-bearing for the claim that the companions will adapt to learners and foster metacognitive growth.
minor comments (3)
  1. [Abstract] The abstract would benefit from briefly indicating the specific educational contexts or levels covered in the five case studies to help readers quickly assess the scope of the illustrations.
  2. Consider adding a short subsection on potential evaluation strategies or metrics for assessing the proposed AI learning companions in future work, to bridge the conceptual framework to practical validation.
  3. Ensure consistent use of terminology when distinguishing AI learning companions from prompted LLM tutors, particularly in the introduction and conclusion.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive feedback on our manuscript. We appreciate the positive assessment of the paper's potential significance for HCI and educational technology. Below, we provide point-by-point responses to the major comments and indicate the revisions we intend to make.

read point-by-point responses
  1. Referee: [Case Studies] Case Studies section: The five case studies review existing tools and note limitations but supply no new controlled data, implementation details, or outcome metrics showing that the proposed pedagogical, adaptive, and responsible foundations actually improve knowledge transfer or learner agency over standard LLM use. This leaves the central claim that the framework resolves the learning-performance paradox unexamined and untested.

    Authors: We acknowledge that the case studies do not include new controlled experiments or quantitative metrics, as the primary contribution of the paper is the proposal of a conceptual framework illustrated by existing tools. The case studies serve to demonstrate the applicability of the three foundations across contexts and to identify gaps in current designs, rather than to empirically validate the framework's superiority. We do not claim that the framework has been tested to resolve the paradox; instead, we argue for the necessity of such a shift and provide a lens for future development. In the revised version, we will clarify this scope in the introduction and conclusion, and add a new subsection on 'Limitations and Future Empirical Directions' to outline how the framework could be tested in subsequent studies. revision: yes

  2. Referee: [Framework] Framework section: The adaptive foundation is outlined at a high level as focusing on how AI learns about students, yet no specific mechanisms (such as student modeling approaches, dynamic feedback loops, or methods for encoding pedagogical principles into LLM behavior) are detailed. This specificity is load-bearing for the claim that the companions will adapt to learners and foster metacognitive growth.

    Authors: The framework is presented at a foundational level to establish guiding principles rather than prescriptive implementations, allowing flexibility for different educational contexts. That said, we agree that additional specificity would strengthen the manuscript. We will revise the adaptive foundation section to include concrete examples of mechanisms, such as the use of Bayesian knowledge tracing or deep knowledge tracing for student modeling, reinforcement learning from human feedback adapted for pedagogical goals, and methods like constitutional AI or supervised fine-tuning to encode pedagogical principles. These additions will illustrate potential pathways without implying that current systems fully realize them. revision: yes

standing simulated objections not resolved
  • The provision of new empirical data or controlled studies demonstrating improved outcomes, since the manuscript is a conceptual proposal and review of existing tools rather than an original empirical investigation.

Circularity Check

0 steps flagged

No circularity: conceptual framework with independent case review

full rationale

The paper advances a position on AI learning companions via three design foundations illustrated by five case studies of existing tools. No equations, fitted parameters, or quantitative predictions appear; claims rest on cited external literature and observed limitations rather than reducing any result to a self-defined input or self-citation chain. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on domain assumptions about how AI affects learning processes and introduces a new conceptual entity without independent falsifiable evidence beyond the proposal itself.

axioms (2)
  • domain assumption Current uses of LLMs in education create a learning-performance paradox that undermines cognitive growth, knowledge transfer, and metacognitive development.
    Invoked in the opening to motivate the need for a new design approach.
  • ad hoc to paper A framework built on pedagogical, adaptive, and responsible foundations can guide the creation of AI systems that prioritize durable learning.
    Core premise of the proposed solution.
invented entities (1)
  • AI learning companions no independent evidence
    purpose: Adaptive, pedagogically informed LLM-powered agents for integration into learning environments.
    New term and concept introduced to distinguish from general-purpose or tutor-prompted LLMs.

pith-pipeline@v0.9.0 · 5800 in / 1525 out tokens · 100201 ms · 2026-05-19T17:34:22.625051+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

119 extracted references · 119 canonical work pages

  1. [1]

    Abdelrahman, Q

    G. Abdelrahman, Q. Wang, and B. Nunes. Knowledge tracing: A survey.ACM Computing Surveys, 55(11):1–37, 2023. 26 Building AI Companions that Prioritise Learning over Performance

  2. [2]

    S. Abdi, H. Khosravi, S. Sadiq, and D. Gasevic. Complementing educational recommender systems with open learner models. InProceedings of the tenth international conference on learning analytics & knowledge, pages 360–365, 2020

  3. [3]

    Aleven, B

    V . Aleven, B. McLaren, I. Roll, and K. Koedinger. Toward tutoring help seeking: Applying cognitive modeling to meta-cognitive skills. InInternational conference on intelligent tutoring systems, pages 227–239. Springer, 2004

  4. [4]

    Aleven, E

    V . Aleven, E. A. McLaughlin, R. A. Glenn, and K. R. Koedinger. Instruction based on adaptive learning technologies. InHandbook of Research on Learning and Instruction, pages 522–560. Routledge, New York, 2 edition, 2016

  5. [5]

    J. R. Anderson, A. T. Corbett, K. R. Koedinger, and R. Pelletier. Cognitive tutors: Lessons learned.The journal of the learning sciences, 4(2):167–207, 1995

  6. [6]

    L. W. Anderson and D. R. Krathwohl.A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom’s Taxonomy of Educational Objectives. Longman, New York, 2001

  7. [7]

    Baker and L

    R. Baker and L. M. Rossi. Assessing the disengaged behaviors of learners.Design recommendations for intelligent tutoring systems, 1:153, 2013

  8. [8]

    R. S. Baker. Stupid tutoring systems, intelligent humans.International journal of artificial intelligence in education, 26(2):600–614, 2016

  9. [9]

    Barnett.A Will to Learn: Being a Student in an Age of Uncertainty

    R. Barnett.A Will to Learn: Being a Student in an Age of Uncertainty. Open University Press, Maidenhead, 2007

  10. [10]

    Bastani, O

    H. Bastani, O. Bastani, A. Sungu, H. Ge, Ö. Kabakcı, and R. Mariman. Generative AI without guardrails can harm learning: Evidence from high school mathematics.Proceedings of the National Academy of Sciences, 122(26):e2422633122, 2025

  11. [11]

    A. Birhane. Algorithmic injustice: A relational ethics approach.Patterns, 2(2):100205, 2021

  12. [12]

    E. L. Bjork and R. A. Bjork. Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In M. A. Gernsbacher and J. R. Pomerantz, editors,Psychology and the Real World: Essays Illustrating Fundamental Contributions to Society, pages 56–64. Worth Publishers, New York, 2011

  13. [13]

    R. A. Bjork, J. Dunlosky, and N. Kornell. Self-regulated learning: Beliefs, techniques, and illusions.Annual Review of Psychology, 64(1):417–444, 2013

  14. [14]

    B. S. Bloom.Taxonomy of Educational Objectives: The Classification of Educational Goals. Longmans Green, New York, 1956

  15. [15]

    Borchers and T

    C. Borchers and T. Shou. Can large language models match tutoring system adaptivity? A benchmarking study. InInternational Conference on Artificial Intelligence in Education, pages 407–420, Cham, 2025. Springer Nature Switzerland

  16. [16]

    J. D. Bransford, A. L. Brown, and R. R. Cocking.How People Learn: Brain, Mind, Experience, and School. National Academy Press, Washington, DC, 2000

  17. [17]

    J. S. Brown, A. Collins, and P. Duguid. Situated cognition and the culture of learning.1989, 18(1):32–42, 1989

  18. [18]

    Brynjolfsson, D

    E. Brynjolfsson, D. Li, and L. Raymond. Generative ai at work.The Quarterly Journal of Economics, 140(2):889– 942, 2025

  19. [19]

    Bulathwela, M

    S. Bulathwela, M. Pérez-Ortiz, C. Holloway, M. Cukurova, and J. Shawe-Taylor. Artificial intelligence alone will not democratise education: On educational inequality, techno-solutionism and inclusive tools.Sustainability, 16(2):781, 2024

  20. [21]

    S. Bull. There are open learner models about!IEEE Transactions on Learning Technologies, 13(2):425–448, 2020

  21. [22]

    Buolamwini and T

    J. Buolamwini and T. Gebru. Gender shades: Intersectional accuracy disparities in commercial gender clas- sification. InProceedings of the 1st Conference on Fairness, Accountability and Transparency, pages 77–91, 2018

  22. [23]

    Cazzaniga, F

    M. Cazzaniga, F. Jaumotte, L. Li, G. Melina, A. J. Panton, C. Pizzinelli, E. J. Rockall, and M. M. Tavares. Gen-AI: Artificial intelligence and the future of work. Staff Discussion Note SDN/2024/001, International Monetary Fund, 2024. 27 Building AI Companions that Prioritise Learning over Performance

  23. [24]

    M. T. Chi and R. Wylie. The icap framework: Linking cognitive engagement to active learning outcomes. Educational psychologist, 49(4):219–243, 2014

  24. [25]

    Chung, M

    J. Chung, M. Henderson, C. Slade, Y . Liang, N. Pepperell, T. Corbin, J. Walton, A. S. Yu, M. Bearman, S. Buckingham Shum, T. Fawns, T. McCluskey, J. McLean, G. Oberg, A. Seligmann, A. Shibani, A. Bakharia, L.-A. Lim, and K. E. Matthews. The use and usefulness of GenAI in higher education: Student experience and perspectives.Computers and Education Open, ...

  25. [26]

    A. T. Corbett and J. R. Anderson. Knowledge tracing: Modeling the acquisition of procedural knowledge.User Modeling and User-Adapted Interaction, 4(4):253–278, 1995

  26. [27]

    B. J. Corbett and J. M. Tangen. Ai tutors vs. tenacious myths: Evidence from personalised dialogue interventions in education.Computers in Human Behavior, 175:108828, 2026

  27. [28]

    Cosyn, H

    E. Cosyn, H. Uzun, C. Doble, and J. Matayoshi. A practical perspective on knowledge space theory: Aleks and its data.Journal of Mathematical Psychology, 101:102512, 2021

  28. [29]

    Darvishi, H

    A. Darvishi, H. Khosravi, S. Sadiq, D. Gaševi ´c, and G. Siemens. Impact of ai assistance on student agency. Computers & Education, 210:104967, 2024

  29. [30]

    Dell’Acqua, E

    F. Dell’Acqua, E. McFowland, E. R. Mollick, H. Lifshitz-Assaf, K. C. Kellogg, S. Rajendran, L. Krayer, F. Candelon, and K. R. Lakhani. Navigating the jagged technological frontier: Field experimental evidence of the effects of Artificial Intelligence on knowledge worker productivity and quality.Organization Science, 2026. First released as Harvard Busines...

  30. [31]

    Dellermann, P

    D. Dellermann, P. Ebel, M. Söllner, and J. M. Leimeister. Hybrid intelligence.Business & Information Systems Engineering, 61(5):637–643, 2019

  31. [32]

    R. Deng, M. Jiang, X. Yu, Y . Lu, and S. Liu. Does chatgpt enhance student learning? a systematic review and meta-analysis of experimental studies.Computers & Education, 227:105224, 2025

  32. [33]

    Denny, S

    P. Denny, S. MacNeil, J. Savelka, L. Porter, and A. Luxton-Reilly. Desirable characteristics for ai teaching assistants in programming education. ITiCSE 2024, page 408–414, New York, NY , USA, 2024. Association for Computing Machinery

  33. [34]

    Dermeval, R

    D. Dermeval, R. Paiva, I. I. Bittencourt, J. Vassileva, and D. Borges. Authoring tools for designing intelligent tutoring systems: A systematic review of the literature.International Journal of Artificial Intelligence in Education, 28(3):336–384, 2018

  34. [35]

    Dillon and colleagues

    E. Dillon and colleagues. Microsoft 365 copilot and knowledge worker productivity: Field experiment. Working paper, 2025. Randomised experiment across 56 firms; placeholder metadata

  35. [36]

    D’Mello and A

    S. D’Mello and A. Graesser. Confusion and its dynamics during device comprehension with breakdown scenarios. Acta Psychologica, 151:196–207, 2014

  36. [37]

    Doroudi, V

    S. Doroudi, V . Aleven, and E. Brunskill. Where’s the reward? a review of reinforcement learning for instructional sequencing.International Journal of Artificial Intelligence in Education, 29(4):568–620, 2019

  37. [38]

    Drachsler and W

    H. Drachsler and W. Greller. Privacy and analytics: it’s a delicate issue—a checklist for trusted learning analytics. InProceedings of the Sixth International Conference on Learning Analytics and Knowledge (LAK 2016), pages 89–98. ACM, 2016

  38. [39]

    Fan et al

    Y . Fan et al. Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance.British Journal of Educational Technology, 56(2):489–530, 2024

  39. [40]

    Floridi, J

    L. Floridi, J. Cowls, M. Beltrametti, R. Chatila, P. Chazerand, V . Dignum, C. Luetge, R. Madelin, U. Pagallo, F. Rossi, B. Schafer, P. Valcke, and E. Vayena. An ethical framework for a good AI society: Opportunities, risks, principles, and recommendations.Minds and Machines, 28(4):689–707, 2018

  40. [41]

    W. Gao, Q. Liu, L. Yue, F. Yao, R. Lv, Z. Zhang, H. Wang, and Z. Huang. Agent4edu: Generating learner response data by generative agents for intelligent education systems. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 23923–23932, 2025

  41. [42]

    Ghosh, N

    A. Ghosh, N. Heffernan, and A. S. Lan. Context-aware attentive knowledge tracing. InProceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, pages 2330–2339, 2020

  42. [43]

    A. C. Graesser, N. K. Person, and J. P. Magliano. Collaborative dialogue patterns in naturalistic one-to-one tutoring.Applied cognitive psychology, 9(6):495–522, 1995

  43. [44]

    X. Han, H. Peng, and M. Liu. The impact of GenAI on learning outcomes: A systematic review and meta-analysis of experimental studies.Educational Research Review, 45:100714, 2025. 28 Building AI Companions that Prioritise Learning over Performance

  44. [45]

    Hardy, S

    J. Hardy, S. P. Bates, M. M. Casey, K. W. Galloway, R. K. Galloway, A. E. Kay, P. Kirsop, and H. A. McQueen. Student-generated content: Enhancing learning through sharing multiple-choice questions.International Journal of Science Education, 36(13):2180–2194, 2014

  45. [46]

    Henderson, M

    M. Henderson, M. Bearman, J. Chung, T. Fawns, S. Buckingham Shum, K. E. Matthews, and J. de Mello Heredia. Comparing generative ai and teacher feedback: student perceptions of usefulness and trustworthiness.Assessment & Evaluation in Higher Education, Online: 13 May 2025:1–16, 2025. doi: 10.1080/02602938.2025.2502582

  46. [47]

    Holmes, K

    W. Holmes, K. Porayska-Pomsta, K. Holstein, E. Sutherland, T. Baker, S. Buckingham Shum, O. C. Santos, M. T. Rodrigo, M. Cukurova, I. I. Bittencourt, et al. Ethics of ai in education: Towards a community-wide framework. International Journal of Artificial Intelligence in Education, 32(3):504–526, 2022

  47. [48]

    M. H. Immordino-Yang and A. Damasio. We feel, therefore we learn: The relevance of affective and social neuroscience to education.Mind, brain, and education, 1(1):3–10, 2007

  48. [49]

    Järvelä, G

    S. Järvelä, G. Zhao, A. Nguyen, and H. Chen. Hybrid intelligence: Human–ai coevolution and learning.British Journal of Educational Technology, 2025

  49. [50]

    F. J.-Y . e. a. Jin. Students’ perceptions of generative ai–powered learning analytics in the feedback process: A feedback literacy perspective.Journal of Learning Analytics, 12(1):152–168, 2025

  50. [51]

    M. Kapur. Examining productive failure, productive success, unproductive failure, and unproductive success in learning.Educational Psychologist, 51(2):289–299, 2016

  51. [52]

    Kazemitabaar, R

    M. Kazemitabaar, R. Ye, X. Wang, A. Z. Henley, P. Denny, M. Craig, and T. Grossman. Codeaid: Evaluating a classroom deployment of an llm-based programming assistant that balances student and educator needs. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, CHI ’24, New York, NY , USA, 2024. Association for Computing Machinery

  52. [53]

    Khosravi, S

    H. Khosravi, S. Buckingham Shum, G. Chen, C. Conati, Y . S. Tsai, J. Kay, S. Knight, R. Martinez-Maldonado, S. Sadiq, and D. Gaševi´c. Explainable artificial intelligence in education.Computers and Education: Artificial Intelligence, 3:100074, 2022

  53. [54]

    Khosravi, G

    H. Khosravi, G. Demartini, S. Sadiq, and D. Gasevic. Charting the design and analytics agenda of learnersourcing systems. InLAK21: 11th international learning analytics and knowledge conference, pages 32–42, 2021

  54. [55]

    Khosravi, P

    H. Khosravi, P. Denny, S. Moore, and J. Stamper. Learnersourcing in the age of ai: Student, educator and machine partnerships for content creation.Computers and Education: Artificial Intelligence, 5:100151, 2023

  55. [56]

    Khosravi, A

    H. Khosravi, A. Shibani, J. Jovanovic, Z. A. Pardos, and L. Yan. Generative ai and learning analytics: Pushing boundaries, preserving principles.Journal of Learning Analytics, 12(1):1–11, 2025

  56. [57]

    Klimecki and T

    O. Klimecki and T. Singer. Empathy from the perspective of social neuroscience.The Cambridge handbook of human affective neuroscience, pages 533–550, 2013

  57. [58]

    K. R. Koedinger, P. F. Carvalho, R. Liu, and E. A. McLaughlin. An astonishing regularity in student learning rate.Proceedings of the National Academy of Sciences, 120(13):e2221311120, 2023

  58. [59]

    K. R. Koedinger, A. Corbett, et al.Cognitive tutors: Technology bringing learning sciences to the classroom. na, 2006

  59. [60]

    D. A. Kolb.Experiential learning: Experience as the source of learning and development. FT press, 2014

  60. [61]

    Kornell and R

    N. Kornell and R. A. Bjork. A stability bias in human memory: Overestimating remembering and underestimating learning.Journal of Experimental Psychology: General, 138(4):449–468, 2009

  61. [62]

    J. W. Lai, W. Qiu, M. Thway, L. Zhang, N. B. Jamil, C. L. Su, S. S. Ng, and F. S. Lim. Leveraging process-action epistemic network analysis to illuminate student self-regulated learning with a socratic chatbot.Journal of Learning Analytics, 12(1):32–49, 2025

  62. [63]

    H.-Y . Lee, J. Kim, H. Choi, H. Bae, A. Jeong, S. Choi, J.-H. Kim, and C.-E. Kim. Comparing ai chatbot simulation and peer role-play for osce preparation: a pilot randomized controlled trial.BMC Medical Education, 25(1):1755, 2025

  63. [64]

    H. Li, T. Xu, C. Zhang, E. Chen, J. Liang, X. Fan, H. Li, J. Tang, and Q. Wen. Bringing generative ai to adaptive learning in education.arXiv preprint arXiv:2402.14601, 2024

  64. [65]

    Liffiton, B

    M. Liffiton, B. E. Sheese, J. Savelka, and P. Denny. Codehelp: Using large language models with guardrails for scalable support in programming classes. InProceedings of the 23rd Koli Calling International Conference on Computing Education Research, Koli Calling ’23, New York, NY , USA, 2024. Association for Computing Machinery. 29 Building AI Companions t...

  65. [66]

    J. Lin, M. Rakovic, D. Lang, D. Gasevic, and G. Chen. Exploring the politeness of instructional strategies from human-human online tutoring dialogues. InLAK22: 12th International Learning Analytics and Knowledge Conference, LAK22, page 282–293, New York, NY , USA, 2022. Association for Computing Machinery

  66. [67]

    W. Ma, O. O. Adesope, J. C. Nesbit, and Q. Liu. Intelligent tutoring systems and learning outcomes: A meta-analysis.Journal of Educational Psychology, 106(4):901–918, 2014

  67. [68]

    Meyer, D

    A. Meyer, D. H. Rose, and D. Gordon.Universal Design for Learning: Theory and Practice. CAST Professional Publishing, Wakefield, MA, 2014

  68. [69]

    Meyer and R

    J. Meyer and R. Land. Threshold concepts and troublesome knowledge: Linkages to ways of thinking and. Princeton: Citeseer, 2003

  69. [70]

    F. Miao, W. Holmes, H. Ronghuai, and Z. Hui.AI and Education: Guidance for Policy-Makers. UNESCO Publishing, Paris, 2021

  70. [71]

    Molenaar

    I. Molenaar. The concept of hybrid human-ai regulation: Exemplifying how to support young learners’ self- regulated learning.Computers and Education: Artificial Intelligence, 3:100070, 2022

  71. [72]

    Molenaar, S

    I. Molenaar, S. de Mooij, R. Azevedo, M. Bannert, S. Järvelä, and D. Gaševi´c. Measuring self-regulated learning and the role of ai: Five years of research using multimodal multichannel data.Computers in Human Behavior, 139:107540, 2023

  72. [73]

    Molenaar, A

    I. Molenaar, A. Horvers, and R. S. Baker. Towards hybrid human-system regulation: Understanding children’srl support needs in blended classrooms. InProceedings of the 9th international conference on learning analytics & knowledge, pages 471–480, 2019

  73. [74]

    Noy and W

    S. Noy and W. Zhang. Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654):187–192, 2023

  74. [75]

    Nugent, J

    A. Nugent, J. M. Lodge, A. Carroll, R. Bagraith, S. MacMahon, K. Matthews, and P. Sah.Higher education learning framework: An evidence informed model for university learning. The University of Queensland, 2019

  75. [76]

    B. D. Nye, A. C. Graesser, and X. Hu. Autotutor and family: A review of 17 years of natural language tutoring. International Journal of Artificial Intelligence in Education, 24(4):427–469, 2014

  76. [77]

    OECD Publishing, Paris, 2026

    OECD.OECD Digital Education Outlook 2026: Exploring Effective Uses of Generative AI in Education. OECD Publishing, Paris, 2026

  77. [78]

    Panadero

    E. Panadero. A review of self-regulated learning: Six models and four directions for research.Frontiers in psychology, 8:422, 2017

  78. [79]

    Panadero and S

    E. Panadero and S. Järvelä. Socially shared regulation of learning: A review.European psychologist, 2015

  79. [80]

    Papamitsiou and A

    Z. Papamitsiou and A. A. Economides. Learning analytics and educational data mining in practice: A systematic literature review of empirical evidence.Journal of educational technology & society, 17(4):49–64, 2014

  80. [81]

    P. I. Pavlik, L. G. Eglington, and L. M. Harrell-Williams. Logistic knowledge tracing: A constrained framework for learner modeling.IEEE Transactions on Learning Technologies, 14(5):624–639, 2021

Showing first 80 references.