1 of 16

Advancing Social Influence Models in Learning Analytic

Joshua M. Rosenberg, University of Tennessee, Knoxville

Bret Staudt Willet, Michigan State University (next, Florida State University)

2 of 16

We are often impacted by our networks

  • Josh’s friend Ryan Estrellado sharing about Ted Lasso on Slack

  • Bret caring about ethical issues in social media research

  • Seeing how colleagues negotiate speaking up in a department meeting

  • TikTok dance competitions1

1We feel very old writing this

3 of 16

These impacts can be seen in terms of influence processes

Influence has long been a key construct in network analysis (Sweet, 2018).

For instance, sociologists developing the social influence approach used statistical models to understand how social capital (i.e., resources inherent to and available through relationships) exerted power (Bourdieu, 1980)

In short, influence may be thought of in terms of how individuals affect one another (Frank, 1998)

Another way to consider influence is in the following terms:

Network -> Individuals’ actions, beliefs, and achieved outcomes

4 of 16

Distinction from another technique - selection

Selection models aim to understand who interacts with whom (Fincham et al., 2018) - these models are different from but related to those we estimate when concerned with influence

These selection processes are contemporarily estimated using powerful extensions of inferential statistical techniques such as logistic regressions, Exponential Random Graph Models (e.g., Gašević et al., 2019)

Another way to consider selection is in the following terms:

Individuals’ actions and beliefs -> A network

5 of 16

Our conjecture

Our central argument here is that influence models are especially valuable because they allow researchers to interrogate what is intuitively important about networks

That is, social networks—those of teachers, administrators, researchers—can influence actions, behaviors, and learning

Using centrality measures as proxies for influential individuals, especially with cross-sectional data, introduce problems with respect to causality (Sweet, 2018)

6 of 16

Examples of influence from others’ work

Frank et al. (2004) examined how the use of innovative digital technologies, namely the use of computers for different purposes, were adopted by teachers throughout a district when teachers identified as leaders among their peers adopted and used them

Collected network data from all of the teachers in the district by asking them to nominate up to ten individuals who they go to for help

Determined how much of the variability in teachers’ use of computer technologies depended upon who they said they went to for help over the preceding year

Found that more variance in computer use was explained by social influence measures—who teachers went to for helpthan the more traditional, psychologically-focused measures of teachers’ value for computers

7 of 16

An example from our own work: #NGSSchat

Goal related to influence: Understand how participation in #NGSSchat conversations across an entire year could explain the rate of participation in the following year

Calculating exposure to others involved determining the number of times every other individual interacted with each individual, and then multiplying that number by in-degree centrality; we then summed these to create a total value-or exposure term-for each individual

In sum, our influence model:

Predicted the number of posts individuals sent in the subsequent year on the basis of an exposure term reflecting their involvement in conversations with central individuals

8 of 16

9 of 16

10 of 16

11 of 16

How the same network could be amenable to selection questions

12 of 16

An example from our own work: #NGSSchat

We found that individuals’ exposure to conversations with central individuals in the previous year was positively associated with greater sustained participation in the next year

Specifically, for every one-SD increase in the number of conversations in which an individual participated, individuals were likely to post 9-15 additional tweets (in log-odds units, β’s = 1.43 – 1.83, p < .001)

In sum, our analysis of Twitter #NGSSchat showed that involvement in conversations (similar to Horn et al., 2020) predicted an outcome—later participation

13 of 16

Future directions

  • Explore under what conditions (and in contexts) networks influence outcomes of interest

    • Carefully considering the nature of outcome, and seek out outcomes beyond the network being studied

  • Drawing from prior research and being theoretically driven in analyses and specifying (and reporting) different means of constructing exposure terms, especially sum or mean of exposure to individuals; weighting by nodal characteristics (e.g., in-degree centrality)

  • Calculating exposure and comparing impacts of exposure from different sources, especially face-to-face and digital networks

14 of 16

Question (and punchline): Should we explore both influence and other aspects of networks?

(we think so, but we’re new to this)

15 of 16

Questions for the field

  • What new questions can influence models allow us to ask?

  • In what ways is influence similar (and complementary) to other network analytic approaches, especially selection model effects through ERGMs?

  • What does limited research using influence models suggest about potential gaps in the growing body of learning analytics research that involves network science?

16 of 16

Contact

Resource: https://datascienceineducation.com/c20.html#c20c