Modeling and Analyzing Collaborative Dialogue States in Mathematics Problem-Solving through ONA and Hidden Markov Models

Abstract

Prior research has shown that collaborative dialogue reflects students’ cognitive, social, and regulatory processes, which support deeper learning. However, less is known about how these moves unfold sequentially during collaborative problem solving (CPS), especially in math contexts. This study leverages Markov Chains, Hidden Markov Models (HMM), and Ordered Network Analysis (ONA) to capture and analyze recurring patterns and transitions in dialogue that are representative of collaboration dynamics and observe how these relate to students’ perceptions of their interactions, collaboration quality, and performance. We conduct our analyses with data from 22 college students (11 dyads) who worked collaboratively to solve algebra problems in a digital learning environment. Results show that dialogue within problems cycled through agreement, clarification, and self-explanation, indicating stable interaction patterns across five HMM latent states. Findings from ONA suggest that while explanatory talk anchors collaboration, task-grounding and co-constructive exchanges may be more directly associated with effective CPS.

Materials

• Full BibTeXapi.crossref.org

Cite

@inproceedings{zhang2026modeling,
  author = {Zhang, Shan and Zambrano, Andrés Felipe and Li, Hongming and Lee, Seiyon and Lee, Ji-Eun and Botelho, Anthony F.},
  title = {Modeling and Analyzing Collaborative Dialogue States in Mathematics Problem-Solving through ONA and Hidden Markov Models},
  booktitle = {Proceedings of the 19th International Conference on Computer-Supported Collaborative Learning - CSCL 2026},
  year = {2026},
  pages = {489--493},
  publisher = {International Society of the Learning Sciences},
  doi = {10.22318/cscl2026.103464},
  url = {https://repository.isls.org/handle/1/12514}
}