AN EXPLORATION OF COLLABORATIVE ACT DISTRIBUTIONS ON THE STACK EXCHANGE NETWORK

ABSTRACT

Social question and answering (social Q&A) services have evolved from digital reference services and expert services, which have connected information seekers with experts or librarians for several decades. Social Q&A services provide the ability to combine the knowledge of many and create an indexable knowledge repository that can be accessed by search engines. They allow users to pose questions and then work with the community to find answers to those questions. Use social Q&A sites have grown immensely in recent years as people turn more frequently to the Internet to find information. Popular services in the United States include Yahoo Answers, Quora, and Stack Exchange. Motivating this research, collaboration in the process of information seeking results in answers that are better than the sum of the individual answers. Understanding the nature of collaboration on social Q&A services can inform design changes that improves the quality of the information that the sites contain.

To investigate the types of collaborations that occur on these sites, a comparison of the collaborative acts occurring on Stack Exchange was conducted, extending the work of Tausczik, Kittur, & Kraut (2014). Stack Exchange exists as a network of identically designed sites federated by topic, with each developing its own community. Coding of existing threads on Stack Exchange, using Stack Overflow, DIY, and Parenting as samples, was completed and the distributions were compared in order to understand the differences in collaboration and provide design recommendations. Findings indicate that the distributions of collaborative acts on Stack Overflow differ significantly from those of DIY and Parenting. Contributions of this research 1) a more inclusive schema of collaborative acts on Stack Exchange, 2) a method for collaborative coding and comparing distributions of collaborative acts, 3) evidence that not all sites on the network share the same distribution, and 4) design recommendations based on the findings.

TABLE OF CONTENTS 

List of Figures ……………………………………………………………………………………………….. vi

List of Tables ………………………………………………………………………………………………… vii

Acknowledgements ………………………………………………………………………………………… viii

Chapter 1 Introduction ……………………………………………………………………………………. 1

Organization of Thesis ……………………………………………………………………………………….. 3

Chapter 2 Literature Review ……………………………………………………………………………. 5

2.1 Q&A Services ……………………………………………………………………………………………… 5

2.1.1 Digital Reference Services …………………………………………………………………… 5

2.1.2 Expert Services …………………………………………………………………………………… 6

2.1.3 Social Q&A ……………………………………………………………………………………….. 7

2.2 Communities of Practice in Social Q&A …………………………………………………………. 11

2.3 CSCW and CIS ……………………………………………………………………………………………. 12

2.3.1 Computer-Supported Cooperative Work ……………………………………………….. 13

2.3.2 Collaborative Information Seeking ……………………………………………………….. 14

Chapter 3 Research Methodology …………………………………………………………………….. 17

3.1 Stack Exchange Description ………………………………………………………………………….. 17

3.2 Research Approach ………………………………………………………………………………………. 20

3.3 Sample Selection ………………………………………………………………………………………….. 22

3.4 Data Collection ……………………………………………………………………………………………. 24

Chapter 4 Findings …………………………………………………………………………………………. 27

4.1 Analysis of 3 Sites ………………………………………………………………………………………… 27

4.2 Pairwise Comparison of 3 Sites ……………………………………………………………………… 29

4.2.1 Comparison of SO and DIY Frequencies ……………………………………………….. 30

4.2.2 Comparison of SO and P Frequencies …………………………………………………… 30

4.2.3 Comparison of DIY and P Frequencies …………………………………………………. 31

Chapter 5 Discussion ……………………………………………………………………………………… 32

5.1 Collaborative Acts on Stack Exchange ……………………………………………………………. 32

5.2 Comparison of Stack Overflow, DIY, and Parenting ………………………………………… 35

5.3 Comparison of Stack Overflow to DIY and Parenting ………………………………………. 37

5.4 Design Recommendations ……………………………………………………………………………… 40

5.5 Contributions ……………………………………………………………………………………………….. 43

5.6 Limitations ………………………………………………………………………………………………….. 45

Chapter 6 Conclusion ……………………………………………………………………………………… 46

6.1 Research Question and Method ……………………………………………………………………… 46

6.2 Contributions and Limitations ……………………………………………………………………….. 47

6.3 Future Work ………………………………………………………………………………………………… 48

Bibliography …………………………………………………………………………………………………. 50

 

 

LIST OF FIGURES

 

 

Figure 1. Four Layer Model Of Information Seeking (Shah, 2009) ……………………………………… 14

Figure 2. A Model For Collaborative Information Seeking (Shah, 2009) ……………………………… 15

Figure 3. Example Of Stack Exchange Question ……………………………………………………………….. 18

Figure 4. Example Of Stack Exchange Answer …………………………………………………………………. 19

Figure 5. Histogram Of Collaborative Act Frequencies ……………………………………………………… 28

 

LIST OF TABLES

 

 

Table 1. Reputation Points and Privileges ………………………………………………………………………… 19

Table 2. Questions Sampled …………………………………………………………………………………………… 23

Table 3. Coding Schema for Each Post ……………………………………………………………………………. 25

Table 4. Distribution of Collaborative Acts ………………………………………………………………………. 27

Table 5. Demographics of communities based on web traffic ……………………………………………… 39

 

Chapter 1 Introduction

While question and answer services have existed before widespread access to the Internet, the web has caused an explosion in the number of users searching for information from the online community. Social question and answering (social Q&A) sites exist as a place for those seeking information to find it using the resources of the online community. Originally, online Q&A was a way to find answers to questions that had a clear solution. Recently, on sites such as Yahoo Answers, Quora, and Stack Exchange, they have evolved into communities of their own, which allow for answering of questions that don’t have such clear answers and allow for a significant amount of collaboration. Much previous research in the area has also shown that Q&A sites are online communities which involve groups of people communicating online, with communities being “long or short term, large or small, national or international, and completely or only partially virtual” (Preece & Maloney-Krichmar, 2003). The primary activities on social Q&A sites fit into the definition of collaborative information seeking, which includes collaborative grounding, or when two or more collaborators “construct shared understanding of the information need and seeking process” (Gazan, 2010). This occurs through comments on questions, comments on answers, up voting, down voting, etc. These collaborations are then made permanently available on the web for others to use as a knowledge base. More information on online communities and collaboration will be provided in the following literature review.

Social Q&A sites typically will have either one forum for questions on any subject or a forum for each specific subject. Stack Exchange exists as a network of sites made with the same boilerplate format, each pertaining to its own subject. As of August 2015, there are 152 different sites in the Stack Exchange network, which are all designed with the exact same format. An advantage of this is that any group that is interested in creating a social Q&A site on a topic can start one.

The way that Stack Exchange is currently designed, there is little consideration for the differences in interactions that might occur between groups that are seeking information or answering questions on different topics. For example, do certain domains require more updating and occur over a longer period of time? Is the new information relevant to the original answerer?

Are users more likely to require clarification on a question or an answer within certain domains? As collaboration is a vital aspect of social Q&A and providing high quality answers to complex questions, developing a deeper understanding of it can help lead to tools that better support it.

From this motivation, the main research question will be answered as well as several subquestions:

Do different Stack Exchange sites lead to different distributions of collaborative acts?

• What are the differences in the distributions of collaborative acts?
• What design changes may be made based on the distribution of collaborative acts that could improve collaboration?

By answering these questions, the researcher seeks to question the efficacy of using the same site design across many, varied domains within Q&A. Designers may use the information found to devise novel interaction mechanisms or better integrate common existing mechanisms for collaboration into each domain’s site. This research will extend the work done by Tausczik, Kittur, and Kraut (2014). In their study of collaboration on MathOverflow, they identified five categories of collaborative acts: provided information, clarified the question, critiqued an answer, revised an answer, and extended an answer. These were used to analyze posts on the website and determine if there was a model that could predict how much the rating of the answer would change followed a collaborative act as well as the frequency of each of the five categories. This study examines several different Stack Exchange sites, including Stack Overflow, the site focusing on programming, Parenting, and Home Improvement. A study of the frequency distributions resulting from this content analysis is then used to call into question qualities of the current Stack Exchange design and provide design recommendations.

Organization of Thesis

This thesis is structured as follows:

• Chapter 2 provides background information on the history of Q&A, describes social Q&A, and discusses related aspects of computer supported cooperative work and collaborative information seeking that provide the theoretical base for this research.
• Chapter 3 describes the Stack Exchange network, research design, the data collection procedure and how rigor was established through multiple initial coders, as well as how the data was analyzed.
• Chapter 4 will report the results of the tests to compare the distributions of collaborative acts across the sites.
• Chapter 5 discusses the findings from this study, including an analysis of the collaborative acts observed, a comparison of the communities, design recommendations, and lastly study contribution and study limitations.
• Chapter 6 summarizes the research questions and method, summarizes contributions and limitations, and provides areas for future work.

AN EXPLORATION OF COLLABORATIVE ACT DISTRIBUTIONS ON THE STACK EXCHANGE NETWORK

Sharing is caring!