1 of 27

Using Platform Data For Social Network Analysis: Cans of Worms, Data Holes, and Magical Thinking

Social Network Analysis Research Team

Virginia Tech

June 22, 2023

2 of 27

Introduction

3 of 27

Introduction - ABCD Model and Social Capital

4 of 27

COVID HAPPENED!

5 of 27

Introduction - Transition to Online

We assumed the strength of faculty network has an effect on the culture of the university, but we could no longer use the same measure of networking.

6 of 27

Introduction - Transition to Online Cont.

Networking was occurring but it was on video- conferencing technology

​

  • “Let’s use the Zoom data. We can anonymize that easily and all we need to know is who is interacting with whom, not what they’re saying. This should be easy and straightforward!”

​

​

​

Famous last words!

​

​

7 of 27

Literature Review

8 of 27

Literature Review - Using SNA to examine communities in Higher Education

“Higher education institutions are woven fabrics, layered and patched together in a complex arrangement. By discerning their structure and the processes by which participants move within, we gain a more intuitive and realistic understanding of how they lived. Nearly every process of education entails relationships and interactions.”

  • (Bianfani and McFarland, 2013, p. 202)

9 of 27

Literature Review - SNA as a Tool to Study Change in Higher Education

Homophily

  • “Social networks are homophilous, meaning that socially linked individuals tend to be similar” - (Lermen et al. 2016, p. 10)

​

  • “Strong ties support the convergence of a community toward implementing best practices effectively but can lead to stagnation and a lack of innovation as the community lacks regular exposure to new ideas from diverse peers” - (Ma et al., 2018)

​

“Birds of a feather flock together”...

but how can we encourage a mix?

​

​

10 of 27

Literature Review - The Importance of Networks in Higher Education

​

  • Regarding the importance of informal relationships for the diffusion of ideas (Kezar, 2014) and decision-making within organizations (Quardokus & Henderson, 2015), change efforts require taking individuals’ relationships seriously.

​

  • Revising teaching practices is often preceded by constant communication and cohesion across groups of STEM faculty. This may be why it takes a long time to change the practices of isolated faculty. - (Ma et al., 2019)

11 of 27

Literature Review - Traditional and Online Forms of SNA in Higher Education

  • Knaub (2018) found that SNA can be used to identify leaders within organizations

​

  • Spalter-Roth (2010) more experienced faculty tend to be closer to the center of faculty networks

​

  • Lane (2010) found that faculty can be influenced by their network to adopt to change.

​

12 of 27

Ideation and IRB

​

13 of 27

Ideation

As we were planning the SNA component of the project, COVID hit, throwing a wrench in our plans…

​

Thought about scrubbing:

  • Social media (i.e. Twitter, Facebook)
  • VT media (i.e. VT emails, VT article comments)

​

What about Zoom?

  • Frequency
  • Length
  • Topic of Meeting

​

14 of 27

Data Request and Collection

Decided upon Snowball Sampling, starting with friends and colleagues that we knew were heavily involved in the Pathways revision

​

Received consent from 17 faculty across the university

​

​

15 of 27

Data Privacy

One of the top priorities was ensuring participant anonymity, which lead us to asking for more general identifiers:

  • College/department of meeting participants
  • Purpose of meeting
  • Frequency of repeating meetings

​

Content of meetings, names of participants, and VT IDs would all remain hidden

​

16 of 27

University Approval and Support

Discussions with TLOS about Data Collection, ensuring anonymity

  • Letter of Approval from Faculty Senate
  • TLOS partnering on the grant (Ognen)

​

Challenges with IRB Approval Process

  • ‘Common Rule’ Revision redefining research
  • Lack of IRB Familiarity with SNA
  • Ensuring anonymity of vulnerable populations

​

​

​

​

​

17 of 27

IRB Process and Timeline

18 of 27

Data Shepherding and Analysis

19 of 27

Challenges with Data Shepherding

  • Incomplete demographic information in TLOS
    • E.g.: student vs. graduate assistant, professor vs. employee

​

  • Complication in data preparations
    • Anonymization
    • Unrecognizable code

​

20 of 27

Data Cleaning Process and Challenges

  • Anonymization, remove duplicities, and etc - Manual efforts
    • Unauthenticated log ins
      • PID vs. alias. vs. preferred email
    • Whether meeting include students

​

  • STATS and analysis – R

​

  • Visualization - UCINET
    • Variable coding – Python

21 of 27

Social Network Analysis

Egocentric Network Analysis -> Snowball Sampling

​

Actors:

  • Ego
  • Alter
  • Tie

​

22 of 27

Ego #7

Spring 2022 Zoom Data with Attributes from Ognen

  • Meeting data: host_ID, participant_ID
  • Attributes: Org unit, Org dept, College, College dept
  • Unknown rate: 56.04%, 84.03%
    • Possible unknown: unauthenticated, gmail, out-of-VT

​

23 of 27

College Data

24 of 27

Department Data

25 of 27

Conclusion and Next Steps

26 of 27

What We Found

What We Thought We Would Find

27 of 27

Next Steps

  • Partner with data stewards and IRB
  • Collect and analyze more Zoom data
  • Triangulate Zoom data with traditional SNA methodologies (e.g. roster)
  • In other words… open more cans of worms!