[Paper Review] Social Interactions Clustering MOOC Students: An Exploratory Study

Lei Shi, Alexandra I. Cristea+3 more|arXiv (Cornell University)|Aug 10, 2020

Online Learning and Analytics10 references4 citations

TL;DR

This study proposes a machine learning approach to cluster MOOC students based on their social interactions in FutureLearn, using comment categorization and statistical modeling. It identifies three stable, distinct student clusters—active contributors, passive responders, and isolated learners—offering insights for improving online course engagement and design.

ABSTRACT

An exploratory study on social interactions of MOOC students in FutureLearn was conducted, to answer "how can we cluster students based on their social interactions?" Comments were categorized based on how students interacted with them, e.g., how a student's comment received replies from peers. Statistical modelling and machine learning were used to analyze comment categorization, resulting in 3 strong and stable clusters.

Motivation & Objective

• To explore how MOOC students can be grouped based on their social interaction behaviors in online forums.
• To identify stable and meaningful clusters of students using comment interaction patterns on FutureLearn.
• To understand the behavioral profiles of learners through statistical modeling of social interactions.
• To support course design and learner engagement strategies by revealing distinct interaction patterns.

Proposed method

• Categorized student comments based on interaction types, such as whether a comment received replies or initiated discussions.
• Applied statistical modeling and machine learning techniques to analyze the structure and patterns in social interactions.
• Used clustering algorithms to group students into distinct behavioral profiles based on interaction frequency and reciprocity.
• Validated cluster stability through repeated analysis and consistency checks across interaction metrics.
• Focused on peer response patterns as a proxy for social engagement, such as number of replies and interaction depth.
• Employed a mixed-methods approach combining quantitative analysis with qualitative interpretation of interaction dynamics.

Experimental results

Research questions

• RQ1 How can MOOC students be meaningfully clustered based on their social interaction behaviors?
• RQ2 What are the dominant patterns of interaction among MOOC learners in a real-world platform like FutureLearn?
• RQ3 Which interaction metrics best distinguish between different types of learner engagement?
• RQ4 How stable and interpretable are the resulting clusters across different analytical approaches?
• RQ5 What behavioral profiles emerge from analyzing peer response and interaction frequency in MOOC forums?

Key findings

• Three stable and distinct clusters of MOOC students were identified: active contributors, passive responders, and isolated learners.
• Active contributors frequently initiated discussions and received many replies, indicating high social engagement.
• Passive responders primarily replied to others’ comments but rarely initiated discussions, showing reactive engagement.
• Isolated learners had minimal interaction, with few comments and almost no replies, suggesting low participation.
• The clustering results were stable across different modeling approaches, indicating robustness of the identified profiles.
• The study demonstrates that social interaction patterns can reliably reveal meaningful learner behavioral types in MOOCs.