Identifying Learners’ Interaction Patterns in an Online Learning Community

Xuemei Wu, Zhenzhen He, Mingxi Li, Zhongmei Han, Changqin HuangInternational Journal of Environmental Research and Public Health 2022 · DOI

TL;DR

This study uses deep learning, social network analysis, and lag sequential analysis to uncover how learners' interaction patterns and emergent roles differ across various co-reflection levels in online learning communities. Understanding these dynamics can help design more effective online learning environments that foster deeper collaborative reflection.

What Problem Does This Paper Address?

Online learning communities offer a rich environment for collaborative reflection, a process crucial for deep knowledge construction. However, it's challenging for learners to achieve high levels of co-reflection, often resorting to superficial contributions. While previous research has touched upon learner roles and co-reflection, a significant gap exists in understanding when and with whom learners at different co-reflection levels interact, and how their emergent roles influence these patterns. This paper aims to bridge this gap by mapping out these complex interaction dynamics.

Methods

The study employs a multi-method approach:

  • Deep Learning (ALBERT & LSTM): To automatically assess learners' co-reflection levels from their textual contributions. This is a novel application for this specific task, leveraging ALBERT for semantic understanding and LSTM for sequential information.
  • Social Network Analysis (SNA): To identify emergent roles (e.g., leader, influencer, peripheral) based on learners' participation, influence, and mediation within the discussion network. This is a standard technique for analyzing group dynamics.
  • Lag Sequential Analysis (LSA): To uncover the temporal interaction patterns between these emergent roles across different co-reflection levels. LSA is used to determine the probability of one role's interaction following another, revealing sequential dynamics.

Datasets

The study utilized data from an online course on fundamental educational technology theory.

  • Training Data: 10,381 co-reflective text posts from similar courses were collected to train the deep learning model.
  • Evaluation Data: 1531 posts from 39 postgraduate students (32 female, 10 male) enrolled in the target course were analyzed. Students who did not participate in any interaction were excluded.

Key Insights and Ideas

  • Inactive Roles, Active Reflection: Most learners, even those with upper-middle co-reflection levels, tend to play inactive roles (regular or peripheral) in the online community. This suggests that passive participation doesn't necessarily equate to low-level reflection.
  • High-Level Learners are Socially Savvy: Learners with high co-reflection levels engage more actively with a diverse range of roles (both peers and teachers) and are more inclined towards self-rethinking.
  • Teacher Influence is Key, but Peer Interaction Matters: Interacting with teachers (guide, facilitator, observer) consistently promotes self-reflection across all co-reflection levels. However, high-level learners initiate interactions with peers before engaging in deep self-reflection, a pattern distinct from lower-level learners.
  • Role Dynamics Drive Interaction: The sequence of interactions with different roles significantly influences co-reflection levels. For instance, high-level learners exhibit more circular interaction patterns, reflecting back and forth between roles, which fosters deeper understanding.

Key Results

  • Co-reflection Levels: Out of 39 learners, 20.51% were low, 48.72% were middle, and 30.77% were high co-reflection level.
  • Emergent Roles: The most common roles were regular (33.33%) and peripheral (30.77%), indicating a tendency towards less active participation. Leader (15.38%), influencer (7.69%), and mediator (7.69%) roles were less frequent, with starters being the rarest (5.13%).
  • Interaction Patterns:
    • Low-level learners: Tend to interact with teachers first, with "guide" and "facilitator" being crucial for prompting self-reflection.
  • Middle-level learners: Engage with influencers more, and self-reflection often occurs after interacting with the observer. They show less independent self-thinking unless prompted.
  • High-level learners: Engage in more bi-directional interactions with teachers and peers, particularly influencers and observers, leading to deeper self-thinking and a more circular reflection process.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Size: The study involved a relatively small sample of 39 postgraduates, limiting the generalizability of the findings.
  • Role Identification: The emergent role identification relied on existing metrics and did not deeply analyze the semantic content of discourse, potentially missing nuances.
  • Interaction Strategy: The study focused on interaction sequences but lacked a detailed analysis of the specific interaction strategies and discourse used.
  • Causality: Explanations for observed patterns are based on existing research and theory, rather than direct causal evidence.

Future work could involve:

  • Larger, Diverse Samples: Recruiting a larger and more diverse participant pool for broader generalizability.
  • Qualitative Analysis: Incorporating qualitative methods like interviews to gain deeper insights into the underlying factors driving interaction patterns and motivations.
  • Advanced Role Positioning: Developing more sophisticated methods for role identification that consider semantic discourse.
  • Scaffolding Design: Designing and testing specific scaffolding interventions based on these findings to enhance learners' co-reflection and examining their effectiveness.

Summary updated

Abstract

The interactions among all members of an online learning community significantly impact collaborative reflection (co-reflection). Although the relationship between learners' roles and co-reflection levels has been explored by previous researchers, it remains unclear when and with whom learners at different co-reflection levels tend to interact. This study adopted multiple methods to examine the interaction patterns of diverse roles among learners with different co-reflection levels based on 11,912 posts. First, the deep learning technique was applied to assess learners' co-reflection levels. Then, a social network analysis (SNA) was conducted to identify the emergent roles of learners. Furthermore, a lag sequence analysis (LSA) was employed to reveal the interaction patterns of the emergent roles among learners with different co-reflection levels. The results showed that most learners in an online learning community reached an upper-middle co-reflection level while playing an inactive role in the co-reflection process. Moreover, higher-level learners were superior in dialog with various roles and were more involved in self-rethinking during the co-reflection process. In particular, they habitually began communication with peers and then with the teacher. Based on these findings, some implications for facilitating online co-reflection from the perspective of roles is also discussed.
BibTeX
@article{wu2022identifying,
  author = {Xuemei Wu and Zhenzhen He and Mingxi Li and Zhongmei Han and Changqin Huang},
  title = {{Identifying Learners’ Interaction Patterns in an Online Learning Community}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2022},
  doi = {10.3390/ijerph19042245},
}
Figure