Learning Analytics for Group Knowledge Building

Bodong Chen, University of Toronto

LAK14 Doctoral Consortium
March 25, 2014

Addressing group-level issues

  • Collaboration
  • Quality of discussion and dialogue
  • New ideas, insights, discoveries, and inventions
  • Morale, socio-cognitive norms, collective cognitive responsibility, civility and other group-level virtues

Knowledge Building

  • Expert teams
  • Sustained, creative work with ideas through communal discourse
    • Cognitive responsibility, epistemic agency
    • Community, discourse
  • One key paradigm in CSCL


Bereiter & Scardamalia

Knowledge Forum

Embedded and transformative assessment

Metadiscourse

“discourse about discourse”

  • How well is our discourse progressing?
  • What progress have we made?
  • Where is our discourse heading?
  • What is hampering progress?
  • Which types of contribution do we need to make?

Metadiscourse in KB

ScaffoldTracker (Chen & Resendes, 2012)

Metadiscourse in KB

Concept Clouds (Resendes, Chen, et al., 2013)

Metadiscourse in KB

Idea Thread Mapper (Zhang, Chen, et al., 2013)

Learning analytics for metadiscourse in group knowledge building?

Trace status of ideas to help better ideas rise


Analytic question

  • How can we reliably assess growth of community knowledge in group knowledge building?
  • Which socio-cognitive factors could we relate this growth to?


Design issue

  • How can we represent the evolution of community knowledge to facilitate metadiscourse to promote knowledge advancement?
  • How to promote metadiscourse conductive to identified benefical factors?

Model growth of community knowledge

  • Latent Dirichlet Allocation (LDA, Shermis & Burstein, 2003)
  • For each contribution, assess its:
    1. semantic topic—the semantic space of topics to which this contribution belongs
    2. novelty—based on semantic comparison with previous contributions, and
    3. quality—based on semantic comparison with pre-established content knowledge benchmarks

Semantic landscape

Evolution of topics

Design tools to promote metadiscourse


  • With understanding of how community knowledge evolves, develop analytic tools to provide a medium for metadiscourse
  • Design classroom interventions to engage students in metadiscourse

Methodology

Reserach approaches

  • Design-based research
  • Co-design

Research sites

Relevant strands

  • Discourse-Centric Analytics
  • Social Learning Analytics
  • Computational Linguistics
  • Knowledge Modelling
  • Statistical Discourse Analysis

General questions

  • Is this work promising?
  • Which direction is (more) promising?
  • Whose work should I build on?
  • How to further articulate the plan?
  • Other considerations?