These images represent a Whole-Thread tie model network analysis of the Edinburgh MOOC forums, using Gephi with a threshold of three except for AI Planning, which uses a threshold of 2.
Those users who did not contribute to the forums were removed to show only those actively participating in this course element, and nodes are sized by betweeness, i.e. the level of connectivity with other nodes.
There are three visualisations for each course: 1. Orange (pale nodes indicate lower eigenvalue centrality and dark red/orange indicate higher values), 2. Green (colourisation by role, from pink (tutor roles) to green (learner roles)), and 3. Multicoloured (colourised by region, using several randomisations of the colour palette to get better contrast).
Colour by region (3) includes a separate colour key for each course. Light pink nodes (2) are Community TAs.
Introduction to Philosophy
The parallel coordinate plot is useful because it shows whether the different courses are similarly placed as the tie model changes, i.e. that we would be likely to draw similar conclusions independent of the tie model. The Whole-Thread tie model without a threshold is most visibly at odds with the general pattern but the relative density values between courses with a threshold of two are similar to it. A tentative conclusion might be that if network structure is sought, using Whole-Thread with a threshold of 2 may be subject to the same “washing-out” problem as without a threshold, even though the absolute value densities are less than 1%.
The Post-Set and Whole-Thread tie models with a threshold are reasonably similar, and with the exception of AI planning, also similar to Poster-Commenter. It might be tentatively suggested that this difference is grounded in different behaviour in the forums; AI Planning participants may have more explicitly commented-on posts.
In summary: at the network level, all of the tie models appear to be acceptable with the exception of Whole-Thread without a threshold. The Poster-Commenter model stands slightly apart from the other tie models and may be usefully discriminating some directed commenting behaviour, consistent with its definition.
This work, unless otherwise expressly stated, is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.