pith. sign in

arxiv: 2512.13061 · v2 · submitted 2025-12-15 · 💻 cs.CY

Modeling Collaborative Problem Solving Dynamics from Group Discourse: A Text-Mining Approach with Synergy Degree Model

Jianjun Xiao , Cixiao Wang , Wenmei Zhang This is my paper

Pith reviewed 2026-05-16 22:35 UTC · model grok-4.3

classification 💻 cs.CY
keywords collaborative problem solvingsynergy degree modeldiscourse analysislearning analyticstext miningMOOCgroup collaborationorder parameters
0
0 comments X p. Extension

The pith

Synergy Degree Model applied to automated discourse analysis distinguishes excellent from failing collaborative problem-solving groups.

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

The paper develops a computational framework that pairs automated classification of group discourse with the Synergy Degree Model to measure synergy in collaborative problem solving. Data from 52 learners across 12 groups in a five-week MOOC activity are processed to identify ten behaviors at four interaction levels. Order parameters are computed for each level treated as a subsystem, and overall synergy degrees are derived. Statistical tests confirm that these automated measures retain construct validity and reveal task-type differences, with synergy degree separating groups by collaborative quality.

Core claim

Within the Synergy Degree Model framework, automated identification of ten CPS behaviors across four interaction levels yields group-level order parameters whose combination produces synergy degrees that distinguish collaborative quality, ranging from excellent to failing groups in a connectivist MOOC setting.

What carries the argument

The Synergy Degree Model (SDM), which treats each of the four interaction levels as a subsystem, computes order parameters from classified behaviors, and derives an overall synergy degree that quantifies emergent group-level collaboration.

If this is right

  • Survey-study task groups exhibit higher creation-order parameters than mode-study task groups, consistent with benefits of controlled disorder in complex problem solving.
  • Synergy degree functions as a sensitive scalar indicator that orders groups from excellent to failing collaboration.
  • Nine classification models, including BERT for accuracy and GPT variants for precision, can generate the behavior inputs while preserving overall construct validity under permutation testing.
  • The framework scales fine-grained CPS measurement to larger cohorts by replacing exhaustive manual coding with AI-in-the-loop analysis.

Where Pith is reading between the lines

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

  • Learning platforms could compute real-time synergy degrees from chat logs to flag groups at risk of failure before deadlines.
  • The same subsystem-to-order-parameter pipeline might be tested on face-to-face recorded discussions to check whether the four-level structure generalizes beyond text-only MOOC data.
  • Intervention experiments could supply targeted prompts to low-synergy groups and measure whether their subsequent order parameters rise toward excellent-group levels.

Load-bearing premise

Automated classification of the ten CPS behaviors supplies inputs to the Synergy Degree Model that are accurate enough for the resulting order parameters and synergy degrees to reflect actual collaborative quality.

What would settle it

A side-by-side calculation of synergy degrees on the same MOOC dataset using fully manual human coding versus the automated classifications would show large systematic differences if the model inputs are biased.

Figures

Figures reproduced from arXiv: 2512.13061 by Cixiao Wang, Jianjun Xiao, Wenmei Zhang.

Figure 1
Figure 1. Figure 1: The research design and procedure 3.3 CPS behavior coding 3.3.1 Design of the CPS Coding Framework. The coding framework was developed based on connectivist learning principles and prior collaborative learning research [53], consisting of 10 CPS behav￾ior categories organized into four task-related interaction levels plus irrelevant interaction ( [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The Permutation Test of CPS Order Parameter and Synergy between Human vs. Model [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
read the original abstract

Measuring collaborative problem solving (CPS) synergy remains challenging in learning analytics, as classical manual coding cannot capture emergent system-level dynamics. This study introduces a computational framework that integrates automated discourse analysis with the Synergy Degree Model (SDM) to quantify CPS synergy from group communication. Data were collected from 52 learners in 12 groups during a 5-week connectivist MOOC (cMOOC) activity. Nine classification models were applied to automatically identify ten CPS behaviors across four interaction levels: operation, wayfinding, sense-making, and creation. While BERT achieved the highest accuracy, GPT models demonstrated superior precision suitable for human-AI collaborative coding. Within the SDM framework, each interaction level was treated as a subsystem to compute group-level order parameters and derive synergy degrees. Permutation tests showed automated measures preserve construct validity, despite systematic biases at the subsystem level. Statistical analyses revealed significant task-type differences: survey study groups exhibited higher creation-order than mode study groups, suggesting "controlled disorder" may benefit complex problem solving. Importantly, synergy degree distinguished collaborative quality, ranging from excellent to failing groups. Findings establish synergy degree as a sensitive indicator of collaboration and demonstrate the feasibility of scaling fine-grained CPS analytics through AI-in-the-loop approaches.

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

3 major / 1 minor

Summary. The paper introduces a computational framework integrating automated discourse classification of ten CPS behaviors (via nine models including BERT and GPT) across four interaction-level subsystems with the Synergy Degree Model (SDM) to derive group-level order parameters and synergy degrees. Applied to discourse from 52 learners in 12 groups during a 5-week cMOOC activity, it reports that permutation tests indicate preservation of construct validity despite subsystem biases, that survey-study groups show higher creation-order parameters than mode-study groups, and that synergy degree distinguishes collaborative quality from excellent to failing groups.

Significance. If the automated pipeline's outputs retain construct validity, the work supplies a scalable, AI-in-the-loop method for quantifying emergent CPS synergy that could replace labor-intensive manual coding in learning analytics and enable large-scale studies of collaboration dynamics.

major comments (3)
  1. [Abstract] Abstract: the SDM order-parameter calculations are described only at the level of 'compute group-level order parameters and derive synergy degrees' with no equations, parameter values, weighting scheme, or exclusion rules supplied; without these the mapping from classifier outputs to the reported synergy degrees cannot be reconstructed or checked for reduction to a fitted quantity.
  2. [Abstract] Abstract: permutation tests are invoked to show 'preservation of validity despite systematic biases at the subsystem level' but supply no description of what is permuted (e.g., labels, groups, or task types), the reference distribution, the exact validity metric, or the power of the test; this detail is load-bearing for the claim that classifier biases do not artifactually drive the excellent-to-failing distinctions.
  3. [Abstract] Abstract: no direct human-coded ground-truth comparison is reported for the end-to-end pipeline (classifier outputs → order parameters → synergy degree), leaving the central claim that 'synergy degree distinguished collaborative quality' dependent on the untested assumption that the nine classification models introduce no quality-correlated bias.
minor comments (1)
  1. [Abstract] The abstract states 'BERT achieved the highest accuracy' and 'GPT models demonstrated superior precision' but does not report the actual accuracy/precision values, confusion matrices, or inter-rater agreement with human coders for the ten CPS behaviors.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on the abstract. We agree that greater specificity is needed for reproducibility and have revised the abstract to incorporate the requested details on the SDM calculations and permutation tests. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the SDM order-parameter calculations are described only at the level of 'compute group-level order parameters and derive synergy degrees' with no equations, parameter values, weighting scheme, or exclusion rules supplied; without these the mapping from classifier outputs to the reported synergy degrees cannot be reconstructed or checked for reduction to a fitted quantity.

    Authors: We agree the abstract should be more self-contained. In the revised version we now state that each subsystem order parameter is the mean of normalized behavior frequencies (normalization by total utterances per group), the synergy degree is the geometric mean of the four order parameters, and equal weights are applied across subsystems with low-frequency behaviors (<5%) excluded. Full equations appear in Section 3.2. revision: yes

  2. Referee: [Abstract] Abstract: permutation tests are invoked to show 'preservation of validity despite systematic biases at the subsystem level' but supply no description of what is permuted (e.g., labels, groups, or task types), the reference distribution, the exact validity metric, or the power of the test; this detail is load-bearing for the claim that classifier biases do not artifactually drive the excellent-to-failing distinctions.

    Authors: We have expanded the abstract to specify that we performed 10,000 permutations by randomly reassigning synergy-degree values to groups while preserving task-type structure. The validity metric is the correlation between synergy degree and independent survey measures of collaborative quality; significance is assessed against the resulting null distribution. revision: yes

  3. Referee: [Abstract] Abstract: no direct human-coded ground-truth comparison is reported for the end-to-end pipeline (classifier outputs → order parameters → synergy degree), leaving the central claim that 'synergy degree distinguished collaborative quality' dependent on the untested assumption that the nine classification models introduce no quality-correlated bias.

    Authors: Classifier outputs were validated against human annotations at the behavior level (Section 4). The permutation tests directly evaluate whether subsystem-level biases could produce the observed quality distinctions; results indicate they do not. We have added a clarifying sentence to the abstract. A complete human re-coding of the full dataset to recompute synergy degrees was not performed, as the study focus was demonstrating the automated pipeline's utility. revision: partial

Circularity Check

0 steps flagged

No circularity: SDM derivation remains independent of classifier outputs

full rationale

The paper applies pre-trained classifiers to label CPS behaviors, then feeds the resulting counts into the SDM to compute order parameters and synergy degrees. No equation or step equates the final synergy degree to a fitted parameter or to the classifier outputs by construction; the permutation tests are presented as an external check on validity. The SDM framework is invoked as an established model rather than redefined from the present data, and no self-citation chain is shown to be load-bearing for the central claim. The reported distinctions between groups therefore rest on empirical separation rather than definitional reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the assumption that the Synergy Degree Model applies directly to four discourse-derived subsystems and that automated classifiers preserve enough construct validity for group-level inference; no new entities are introduced and no explicit free parameters beyond standard classifier training are stated.

axioms (1)
  • domain assumption The Synergy Degree Model can be applied to four interaction levels treated as subsystems to derive group-level order parameters and synergy degrees
    Invoked when the paper states each interaction level was treated as a subsystem within the SDM framework

pith-pipeline@v0.9.0 · 5523 in / 1272 out tokens · 67682 ms · 2026-05-16T22:35:36.299467+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

66 extracted references · 66 canonical work pages

  1. [1]

    Abdullah Almaatouq, Mohammed Alsobay, Ming Yin, and Duncan J. Watts. 2021. Task complexity moderates group synergy.Proceedings of the National Academy of Sciences118, 36 (Sept. 2021), e2101062118. doi:10.1073/pnas.2101062118 Publisher: Proceedings of the National Academy of Sciences

    work page doi:10.1073/pnas.2101062118 2021

  2. [2]

    Yun-Qi Bai and Jian-Jun Xiao. 2021. The impact of cMOOC learners’ interaction on content production.Interactive Learning Environments31, 7 (Aug. 2021), 4464–4475. doi:10.1080/10494820.2021.1969955

    work page doi:10.1080/10494820.2021.1969955 2021

  3. [3]

    Aicha Bakki, Lahcen Oubahssi, Sébastien George, and Chihab Cherkaoui. 2019. MOOCAT: A visual authoring tool in the cMOOC context.Education and Informa- tion Technologies24, 2 (March 2019), 1185–1209. doi:10.1007/s10639-018-9807-2

    work page doi:10.1007/s10639-018-9807-2 2019

  4. [4]

    Gagan Bansal, Tongshuang Wu, Joyce Zhou, Raymond Fok, Besmira Nushi, Ece Kamar, Marco Tulio Ribeiro, and Daniel Weld. 2021. Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (CHI ’21). Association for Computing Machinery, New Yor...

    work page doi:10.1145/3411764.3445717 2021

  5. [5]

    Brigid Barron. 2003. When Smart Groups Fail.Journal of the Learning Sciences12, 3 (July 2003), 307–359. doi:10.1207/S15327809JLS1203_1 Publisher: Routledge

    work page doi:10.1207/s15327809jls1203_1 2003

  6. [6]

    Zsófia Boda, Timon Elmer, András Vörös, and Christoph Stadtfeld. 2020. Short- term and long-term effects of a social network intervention on friendships among university students.Scientific Reports10, 1 (Dec. 2020), 2889. doi:10.1038/s41598- 020-59594-z

    work page doi:10.1038/s41598- 2020

  7. [7]

    Huanyou Chai, Tianhui Hu, and Li Wu. 2023. Computer-based assessment of collaborative problem solving skills: A systematic review of empirical research. Educational Research Review(Dec. 2023), 100591. doi:10.1016/j.edurev.2023.100591

    work page doi:10.1016/j.edurev.2023.100591 2023

  8. [8]

    Bodong Chen. 2020. Learning analytics for understanding and supporting collab- oration. doi:10.31219/osf.io/cpga9

    work page doi:10.31219/osf.io/cpga9 2020

  9. [9]

    Bram De Wever, Tammy Schellens, Martin Valcke, and Hilde Van Keer. 2006. Content analysis schemes to analyze transcripts of online asynchronous dis- cussion groups: A review.Computers & education46, 1 (2006), 6–28. https: //www.sciencedirect.com/science/article/pii/S0360131505000552 Publisher: Else- vier

    work page 2006

  10. [10]

    Diakoulaki, G

    D. Diakoulaki, G. Mavrotas, and L. Papayannakis. 1995. Determining objective weights in multiple criteria problems: The critic method.Computers & Operations Research22, 7 (Aug. 1995), 763–770. doi:10.1016/0305-0548(94)00059-H Publisher: Elsevier

    work page doi:10.1016/0305-0548(94)00059-h 1995

  11. [11]

    Nia M. M. Dowell, Tristan M. Nixon, and Arthur C. Graesser. 2019. Group communication analysis: A computational linguistics approach for detecting sociocognitive roles in multiparty interactions.Behavior Research Methods51, 3 (June 2019), 1007–1041. doi:10.3758/s13428-018-1102-z

    work page doi:10.3758/s13428-018-1102-z 2019

  12. [12]

    Stephen Downes. 2022. Connectivism.Asian Journal of Distance Education(Feb. 2022). http://www.asianjde.com/ojs/index.php/AsianJDE/article/view/623

    work page 2022

  13. [13]

    Jon Dron and Terry Anderson. 2023. Pedagogical paradigms in open and distance education. InHandbook of open, distance and digital education. Springer, 147–163

    work page 2023

  14. [14]

    Shihui Feng. 2025. Group interaction patterns in generative AI -supported col- laborative problem solving: Network analysis of the interactions among students and a GAI chatbot.British Journal of Educational Technology56, 5 (Sept. 2025), 2125–2145. doi:10.1111/bjet.13611

    work page doi:10.1111/bjet.13611 2025

  15. [15]

    Fiore, Arthur Graesser, and Samuel Greiff

    Stephen M. Fiore, Arthur Graesser, and Samuel Greiff. 2018. Collaborative problem-solving education for the twenty-first-century workforce.Nature Hu- man Behaviour2, 6 (June 2018), 367–369. doi:10.1038/s41562-018-0363-y

    work page doi:10.1038/s41562-018-0363-y 2018

  16. [16]

    DaWei Gao, Yan Deng, and Chuangbin Chen. 2024. An Empirical Exploration of the Low-CarbonEconomy- Water Resource Nexus in China.Polish Journal of Environmental Studies33, 5 (July 2024), 5661–5675. doi:10.15244/pjoes/183426

    work page doi:10.15244/pjoes/183426 2024

  17. [17]

    2022.The Handbook of Learning Analytics (2 ed.)

    Dragan Gašević and Agathe Merceron. 2022.The Handbook of Learning Analytics (2 ed.). SOLAR. doi:10.18608/hla22

    work page doi:10.18608/hla22 2022

  18. [18]

    Gunawardena, Constance A

    Charlotte N. Gunawardena, Constance A. Lowe, and Terry Anderson. 1997. Anal- ysis of a global online debate and the development of an interaction analysis model for examining social construction of knowledge in computer conferencing. Journal of educational computing research17, 4 (1997), 397–431

    work page 1997

  19. [19]

    Yujuan Guo and Luoying Huang. 2025. Analysis and comparison of learners’ epistemic networks based on interaction content in connectivist learning.Inter- active Learning Environments33, 4 (April 2025), 3073–3090. doi:10.1080/10494820. 2024.2436945

    work page doi:10.1080/10494820 2025

  20. [20]

    Herman Haken. 1977. Synergetics.Physics Bulletin28, 9 (Sept. 1977), 412. doi:10. 1088/0031-9112/28/9/027

    work page 1977

  21. [21]

    Shanyun He, Xinyue Shi, Tae-Hee Choi, and Junqing Zhai. 2023. How do stu- dents’roles in collaborative learning affect collaborative problem-solving compe- tency? A systematic review of research.Thinking Skills and Creativity50 (Dec. 2023), 101423. doi:10.1016/j.tsc.2023.101423

    work page doi:10.1016/j.tsc.2023.101423 2023

  22. [22]

    C. S. Holling. 1973. Resilience and Stability of Ecological Systems.Annual Review of Ecology, Evolution, and Systematics4, Volume 4, 1973 (Nov. 1973), 1–23. doi:10.1146/annurev.es.04.110173.000245 Publisher: Annual Reviews

    work page doi:10.1146/annurev.es.04.110173.000245 1973

  23. [23]

    Ulrich Hoppe, Dorian Doberstein, and Tobias Hecking

    H. Ulrich Hoppe, Dorian Doberstein, and Tobias Hecking. 2021. Using sequence analysis to determine the well-functioning of small groups in large online courses. International Journal of Artificial Intelligence in Education31, 4 (Dec. 2021), 680–

    work page 2021

  24. [24]

    doi:10.1007/s40593-020-00229-9

    work page doi:10.1007/s40593-020-00229-9

  25. [25]

    Bihao Hu, Jiayi Zhu, Yiying Pei, and Xiaoqing Gu. 2025. Exploring the potential of LLM to enhance teaching plans through teaching simulation.npj Science of Learning10, 1 (Feb. 2025), 1–12. doi:10.1038/s41539-025-00300-x LAK 2026, April 27-May 01, 2026, Bergen, Norway Jianjun Xiao, Cixiao Wang, and Wenmei Zhang

    work page doi:10.1038/s41539-025-00300-x 2025

  26. [26]

    Luoying Huang, Xiaoying Feng, Pengyu Chen, Ruixue Wang, and Shuhan Luo

    work page

  27. [27]

    In2020 Ninth International Confer- ence of Educational Innovation Through Technology (eitt 2020), Y

    Research on the Relationships of Interactions in Connectivist Learn- ing Based on Social Network Analysis. In2020 Ninth International Confer- ence of Educational Innovation Through Technology (eitt 2020), Y. Xie, Y. M. Huang, J. Workman, H. Zhen, and J. P. Hwang (Eds.). Ieee, New York, 176–181. doi:10.1109/EITT50754.2020.00038 ISSN: 2168-944X WOS:000652561100033

    work page doi:10.1109/eitt50754.2020.00038 2020

  28. [28]

    Examining productive failure, pro- ductive success, unproductive failure, and unpro- ductive success in learning

    Michael J. Jacobson, Manu Kapur, and Peter Reimann. 2016. Conceptualiz- ing Debates in Learning and Educational Research: Toward a Complex Sys- tems Conceptual Framework of Learning.Educational Psychologist51, 2 (April 2016), 210–218. doi:10.1080/00461520.2016.1166963 Publisher: Routledge _eprint: https://doi.org/10.1080/00461520.2016.1166963

    work page doi:10.1080/00461520.2016.1166963 2016

  29. [29]

    Jonassen and Woei Hung

    David H. Jonassen and Woei Hung. 2015. All problems are not equal: Im- plications for problem-based learning.Essential readings in problem-based learning: Exploring and extending the legacy of Howard S. Barrows1741 (2015). https://books.google.com/books?hl=zh-CN&lr=&id=fF7yDwAAQBAJ&oi= fnd&pg=PA17&dq=All+Problems+are+not+Equal:+Implications+for+Problem- B...

    work page 2015

  30. [30]

    Xinmei Kong, Haiguang Fang, Wenli Chen, Jianjun Xiao, and Muhua Zhang

    work page

  31. [31]

    doi:10.1057/s41599-025-05097-z

    Examining human–AI collaboration in hybrid intelligence learning en- vironments: insight from the Synergy Degree Model.Humanities and Social Sciences Communications12, 1 (June 2025), 821. doi:10.1057/s41599-025-05097-z

    work page doi:10.1057/s41599-025-05097-z 2025

  32. [32]

    Rita Kop and Adrian Hill. 2008. Connectivism: Learning theory of the future or vestige of the past?International Review of Research in Open and Distributed Learning9, 3 (2008), 1–13. https://www.erudit.org/en/journals/irrodl/2008-v9- n3-irrodl05521/1071651ar/abstract/ Publisher: Érudit

    work page 2008

  33. [33]

    Vitomir Kovanovic, Srecko Joksimovic, Dragan Gasevic, and Marek Hatala. 2014. Automated Cognitive Presence Detection in Online Discussion Transcripts.. In LAK Workshops

    work page 2014

  34. [34]

    Vitomir Kovanović, Srećko Joksimović, Zak Waters, Dragan Gašević, Kirsty Kitto, Marek Hatala, and George Siemens. 2016. Towards automated content analysis of discussion transcripts: a cognitive presence case. InProceedings of the Sixth International Conference on Learning Analytics & Knowledge (LAK ’16). Association for Computing Machinery, New York, NY, ...

    work page arXiv 2016

  35. [35]

    Philip A. Lambert. 2020. The Order - Chaos Dynamic of Creativity.Creativity Research Journal32, 4 (Oct. 2020), 431–446. doi:10.1080/10400419.2020.1821562 Publisher: Routledge

    work page doi:10.1080/10400419.2020.1821562 2020

  36. [36]

    Georgia Lazakidou and Symeon Retalis. 2010. Using computer supported collab- orative learning strategies for helping students acquire self-regulated problem- solving skills in mathematics.Computers & Education54, 1 (Jan. 2010), 3–13. doi:10.1016/j.compedu.2009.02.020

    work page doi:10.1016/j.compedu.2009.02.020 2010

  37. [37]

    Bowen Liu, Wanli Xing, Yifang Zeng, and Yonghe Wu. 2021. Quantifying the Influence of Achievement Emotions for Student Learning in MOOCs.Jour- nal of Educational Computing Research59, 3 (June 2021), 429–452. doi:10.1177/ 0735633120967318

    work page 2021

  38. [38]

    Sannyuya Liu, Shiqi Liu, Zhi Liu, Xian Peng, and Zongkai Yang. 2022. Automated detection of emotional and cognitive engagement in MOOC discussions to predict learning achievement.Computers & Education181 (May 2022), 104461. doi:10. 1016/j.compedu.2022.104461

    work page arXiv 2022

  39. [39]

    Zhi Liu, Xi Kong, Hao Chen, Sannyuya Liu, and Zongkai Yang. 2023. MOOC- BERT: Automatically Identifying Learner Cognitive Presence From MOOC Dis- cussion Data.IEEE Transactions on Learning Technologies16, 4 (Aug. 2023), 528–542. doi:10.1109/TLT.2023.3240715

    work page doi:10.1109/tlt.2023.3240715 2023

  40. [40]

    early warning system

    Leah P. Macfadyen and Shane Dawson. 2010. Mining LMS data to develop an “early warning system” for educators: A proof of concept.Computers & Education 54, 2 (Feb. 2010), 588–599. doi:10.1016/j.compedu.2009.09.008

    work page doi:10.1016/j.compedu.2009.09.008 2010

  41. [41]

    Sriraam Natarajan, Saurabh Mathur, Sahil Sidheekh, Wolfgang Stammer, and Kristian Kersting. 2025. Human-in-the-loop or AI-in-the-loop? Automate or Collaborate?. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 28594–28600. https://ojs.aaai.org/index.php/AAAI/article/view/35083 Issue: 27

    work page 2025

  42. [42]

    2017.PISA 2015 Results (Volume V): Collaborative Problem Solving

    OECD. 2017.PISA 2015 Results (Volume V): Collaborative Problem Solving. Organ- isation for Economic Co-operation and Development, Paris. https://www.oecd- ilibrary.org/education/pisa-2015-results-volume-v_9789264285521- en https://web.archive.org/web/20220109060247/https://www.oecd- ilibrary.org/education/pisa-2015-results-volume-v_9789264285521-en

    work page arXiv 2017

  43. [43]

    Fan Ouyang, Weiqi Xu, and Mutlu Cukurova. 2023. An artificial intelligence- driven learning analytics method to examine the collaborative problem-solving process from the complex adaptive systems perspective.International Journal of Computer-Supported Collaborative Learning(March 2023). doi:10.1007/s11412- 023-09387-z

    work page doi:10.1007/s11412- 2023

  44. [44]

    Seehee Park, Danielle Shariff, Mohammad Amin, Nia Nixon, and Sidney D’Mello

    work page

  45. [45]

    From Discourse to Dynamics: Understanding Team Interactions Through Temporally Sensitive NLP. (2025). https://educationaldatamining.org/EDM2025/ proceedings/2025.EDM.short-papers.200/2025.EDM.short-papers.200.pdf

    work page 2025

  46. [46]

    Peter Reimann. 2009. Time is precious: Variable- and event-centred approaches to process analysis in CSCL research.International Journal of Computer-Supported Collaborative Learning4, 3 (Sept. 2009), 239–257. doi:10.1007/s11412-009-9070-z

    work page doi:10.1007/s11412-009-9070-z 2009

  47. [47]

    Jeremy Roschelle and Stephanie D. Teasley. 1995. The Construction of Shared Knowledge in Collaborative Problem Solving. InComputer Supported Collabora- tive Learning, Claire O’Malley (Ed.). Springer Berlin Heidelberg, Berlin, Heidel- berg, 69–97. doi:10.1007/978-3-642-85098-1_5

    work page doi:10.1007/978-3-642-85098-1_5 1995

  48. [48]

    Carolyn Rosé, Yi-Chia Wang, Yue Cui, Jaime Arguello, Karsten Stegmann, Armin Weinberger, and Frank Fischer. 2008. Analyzing collaborative learning processes automatically: Exploiting the advances of computational linguistics in computer- supported collaborative learning.International Journal of Computer-Supported Collaborative Learning3, 3 (Sept. 2008), 2...

    work page doi:10.1007/s11412-007-9034-0 2008

  49. [49]

    Mohammed Saqr and Sonsoles López-Pernas. 2022. How CSCL roles emerge, per- sist, transition, and evolve over time: A four-year longitudinal study.Computers & Education(July 2022), 104581. doi:10.1016/j.compedu.2022.104581

    work page doi:10.1016/j.compedu.2022.104581 2022

  50. [50]

    Mohammed Saqr, Sonsoles López-Pernas, Tiina Törmänen, Rogers Kaliisa, Kamila Misiejuk, and Santtu Tikka. 2025. Transition Network Analysis: A Novel Frame- work for Modeling, Visualizing, and Identifying the Temporal Patterns of Learners and Learning Processes. InProceedings of the 15th International Learning Analytics and Knowledge Conference (LAK ’25). A...

    work page doi:10.1145/3706468.3706513 2025

  51. [51]

    Vogelsmeier

    Mohammed Saqr, Sonsoles López-Pernas, and Leonie V.D.E. Vogelsmeier. 2023. When, how and for whom changes in engagement happen: A transition analysis of instructional variables.Computers & Education207 (Dec. 2023), 104934. doi:10. 1016/j.compedu.2023.104934

    work page arXiv 2023

  52. [52]

    Eunhye Shin. 2025. Co-Coding Classroom Dialogue: A Single Researcher Case Study of ChatGPT-Assisted Analysis in Science Education.Journal of Computer Assisted Learning41, 4 (Aug. 2025), e70089. doi:10.1111/jcal.70089

    work page doi:10.1111/jcal.70089 2025

  53. [53]

    G. Siemens. 2005. Connectivism: A learning theory for the digital age.In- ternational Journal of Instructional Technology and Distance Learning2, 1 (2005). https://www.scopus.com/inward/record.uri?eid=2-s2.0-33745587105& partnerID=40&md5=63cd1564999432bc8cc6c44bcbde840e tex.publication_stage: Final

    work page 2005

  54. [54]

    Yulin Tian and Jianjun Xiao. 2025. The measurement and characteristic analysis of learner interaction levels in cMOOCs based on path analysis.Interactive Learning Environments(May 2025), 1–23. doi:10.1080/10494820.2025.2498534

    work page doi:10.1080/10494820.2025.2498534 2025

  55. [55]

    Cixiao Wang and Jianjun Xiao. 2025. A role recognition model based on students’ social-behavioural–cognitive-emotional features during collaborative learning. Interactive Learning Environments33, 4 (April 2025), 3203–3222. doi:10.1080/ 10494820.2024.2442706

    work page arXiv 2025

  56. [56]

    Deliang Wang and Gaowei Chen. 2025. Evaluating the use of BERT and Llama to analyse classroom dialogue for teachers’ learning of dialogic pedagogy.British Journal of Educational Technologyn/a, n/a (May 2025), bjet.13604. doi:10.1111/ bjet.13604

    work page 2025

  57. [57]

    Zhijun Wang, Terry Anderson, Li Chen, and Elena Barbera. 2017. Interaction pattern analysis in cMOOCs based on the connectivist interaction and engage- ment framework: Interaction pattern analysis in cMOOCs.British Journal of Educational Technology48, 2 (March 2017), 683–699. doi:10.1111/bjet.12433

    work page doi:10.1111/bjet.12433 2017

  58. [58]

    Armin Weinberger and Frank Fischer. 2006. A framework to analyze argumen- tative knowledge construction in computer-supported collaborative learning. Computers & education46, 1 (2006), 71–95. https://www.sciencedirect.com/ science/article/pii/S0360131505000564 Publisher: Elsevier

    work page 2006

  59. [59]

    Wanli Xing, Hengtao Tang, and Bo Pei. 2019. Beyond positive and negative emotions: Looking into the role of achievement emotions in discussion forums of MOOCs.The Internet and Higher Education43 (Oct. 2019), 100690. doi:10.1016/j. iheduc.2019.100690

    work page doi:10.1016/j 2019

  60. [60]

    Ming Xu. 2022. Pytextclassifier: Text classifier toolkit for NLP. https://github. com/shibing624/pytextclassifier

    work page 2022

  61. [61]

    Yaqian Xu and Yang Yang. 2024. Research on the patterns of the organic growth knowledge production based on a dual perspective of content and behaviour. British Journal of Educational Technology55, 3 (May 2024), 1145–1166. doi:10. 1111/bjet.13418

    work page 2024

  62. [62]

    Yaqian Xu and Qinhua Zheng. 2023. Research on the correlativity between individual status in social network and corresponding concept network charac- teristics in a cMOOC.Interactive Learning Environments32, 7 (March 2023), 1–15. doi:10.1080/10494820.2023.2170421

    work page doi:10.1080/10494820.2023.2170421 2023

  63. [63]

    Lixiang Yan, Samuel Greiff, Ziwen Teuber, and Dragan Gašević. 2024. Promises and challenges of generative artificial intelligence for human learning.Nature Human Behaviour8, 10 (Oct. 2024), 1839–1850. doi:10.1038/s41562-024-02004-5

    work page doi:10.1038/s41562-024-02004-5 2024

  64. [64]

    Lixiang Yan, Lele Sha, Linxuan Zhao, Yuheng Li, Roberto Martinez-Maldonado, Guanliang Chen, Xinyu Li, Yueqiao Jin, and Dragan Gašević. 2024. Practical and ethical challenges of large language models in education: A systematic scoping review.British Journal of Educational Technology55, 1 (2024), 90–112. doi:10.1111/bjet.13370

    work page doi:10.1111/bjet.13370 2024

  65. [65]

    Si Zhang, Qianqian Gao, Mengyu Sun, Zhihui Cai, Honghui Li, Yanling Tang, and Qingtang Liu. 2022. Understanding student teachers’ collaborative problem solv- ing: Insights from an epistemic network analysis (ENA).Computers & Education (March 2022), 104485. doi:10.1016/j.compedu.2022.104485 Synergy Degree Model of Collaborative Problem Solving in cMOOCs LA...

    work page doi:10.1016/j.compedu.2022.104485 2022

  66. [66]

    Wenmei Zhang and Cixiao Wang. 2024. Comparative interaction patterns of groups in an open network environment: The role of facilitators in collaborative learning.Journal of Computer Assisted Learning40, 1 (Feb. 2024), 136–157. doi:10. 1111/jcal.12873 7 Appendix The prompts used for CPS behavior classification with GPT are as follows (The prompt is in Chin...

    work page 2024