Title: Identity-Based Signals and E-Mentorship to Support Engagement in Online Programming Communities

Examination Committee:

Dr. Christopher Parnin (Chair)
Dr. Tiffany Barnes (Member)
Dr. James Herbsleb (External Member)
Dr. Christopher Mayhorn (Graduate School Representative), (Cognitive Science Minor Representative)
Dr. Kathryn Stolee (Member)

All members of the university community are invited.

* * *
Abstract: Online programming communities, a type of socio-technical ecosystem, enable a peer support cycle where software developers can receive technical guidance from other developers who encounter similar issues. With over 23 million software developers in the world, these communities are helpful for gaining a global perspective on a range of complex programming problems. Communities such as Stack Overflow and GitHub have a global reach of over 10 million and 37 million registered users, respectively. These numbers do not take into account the large number of users without registered accounts—who are able to derive a high utility from these platforms.

Despite the many advantages of engaging in these communities, developers who identify as novices and underrepresented groups do not. According to global annual developer surveys, women are less than 8% of the community and less than 2% are non-binary. Similar distributions occur when referring to race and ethnicity: the majority of contributors to these communities are White or of European descent (70%). Likewise, novice programmers—forming a third of the community with less than 5 years of experience—continue to lurk and express interest in participating but face challenges finding opportunities to contribute. The existing mechanisms in place to engage these users often focus on getting developers to the platform or creating a new one altogether. But how can we invite developers to engage in a community that is already broken? My approach focuses on understanding and building new community mechanisms to create a more inclusive online programming community.

The goal of this research is to investigate inhibitors to engagement in online programming communities using the conceptual framework of barriers. The conceptual framework of barriers are split into two groups: social barriers—obstacles that limit the social interactions users seek and cognitive barriers—obstacles that lead to cognitive burdens users with endeavors to engage in the community encounter. Using this framework as a roadmap of challenges users face, we can derive interventions to support engagement.

The thesis of this dissertation is that existing mechanisms in online programming communities do not make the contribution process inclusive for novice and underrepresented programmers due to existing social and cognitive barriers. By incorporating identity-based signals, introducing mentorship, and understanding sub-communities, we can help programmers overcome these barriers and significantly increase participation in online programming communities. Through the conceptual framework of barriers, this dissertation understands and build mechanisms of how barriers can be reduced and engagement increased in four studies:

1. To learn how identity can encourage engagement, I conducted an empirical analysis of post and user data on Stack Overflow of women helping other women. I found that only 32% of identifiable women have ever posted a question. I also found that women who experience peer parity were more likely to engage sooner.

2. To learn how e-mentoring can influence engagement, I crafted a collaborative, formative feedback forum on live Stack Overflow. I find that with just-in-time mentoring, we can reduce negative experiences for participants and improve community receptiveness to novice contributions. I found that mentored questions provided feedback that improved novice’s question quality. I found that the average question score increased over 50%, and novices were extremely satisfied with their mentorship experience.

3. To learn how barriers are influenced by sub-communities, I conducted a study immersing software developers in a Stack Overflow Team instance while they worked on their software projects and monitored their activity. I found that while most developers did not ask questions, their perspective of these barriers inhibiting their participation was influenced. I also found that having an offline relationship with online community members increased trust and belonging to the community.

4. To learn how identity signals are used in contributing code in online programming communities, I designed an eye-tracking experiment with programmers as they reviewed pull requests. I examined if and how supplementary technical details such as previous contributions, and socially identifying connections such as the avatar image, are used when making decisions about code contributions. I found that even when they do not think they are, programmers consider social signals of users when asked to review code contributions.

This dissertation concludes with implications for this work and future directions that lead towards explicit mechanisms to transfer offline acquaintances into online sustained bonds, create systems that encourage guidance or e-mentoring with close proximity to a shared identity, and building community infrastructure support sub-community and micro-community engagement.
—
Computer Science Graduate Programs
go.ncsu.edu/csc-grad-resources
NC State University