|
Article Excerpt
1. Introduction
In recent years we have witnessed an immense proliferation of open source software (1) (hereafter, OSS) development and a radical shift in paradigm from the traditional software-engineering practice. At the center of this burgeoning phenomenon are the participants--project leaders, developers, and users--who provide voluntary "community service" (i.e., software development/improvement, bug fixes, peer reviews, etc.) in return for nothing but intrinsic rewards (von Hippel and von Krogh 2003). These participants, who are scattered around the globe, form virtual communities in which dynamic interactions take place in the pursuit of common goals (von Krogh et al. 2003, Jones et al. 2004).
Interestingly, however, not all open source projects appear to be active, nor do they all receive the same level of attention and support as major open source initiatives such as Linux, arguably the most well-known OSS. For example, despite its enormous potential benefits, Hypermail, once recognized as a highly promising open source e-mail application, failed to attract and maintain the participants necessary to enhance the functionalities and features of the application. This particular community has been very silent over the past few years, and no significant progress has recently been made.
Recently, the stability of many OSS communities has been severely threatened by the presence of various "external influences," such as the availability of other OSS projects, lucrative financial incentives, and career offers by established firms (Lerner and Tirole 2002), and commercial vendor intervention (O'Mahony 2003), all of which can cause a change in the attitude of participants and thus influence membership dynamics. Because OSS participants provide services without formal contracts, they can freely leave the community, just as they can freely join. This danger is always present in any OSS community (Markus et al. 2000, Butler 2001).
In light of these issues, questions arise: (1) How does a participant's membership decision to stay or leave (2) affect the other individuals ("neighbors") with whom he or she has collaborated over an extended period of time within an OSS network? (3) (2) To what extent do network characteristics (i.e., size and connectivity) mediate the impact of these external factors on the OSS participants' dynamic membership decisions, and hence the stability of the network? This paper addresses these two questions with an emphasis on membership dynamics. Whereas the first question is addressed theoretically from the Ising (1925) perspective, the second is addressed by performing simulations with the bounds of empirical network data that were collected from two actual OSS communities, Linux and Hypermail.
We consider the Ising perspective a suitable mechanism for exploring OSS network dynamics. The Ising theory is generally concerned with basic patterns in dynamic interactions among physical objects or economic agents. A computer simulation was carefully designed to validate our theoretical propositions. Based on the results obtained from the simulation, we explore the implications for OSS communities with respect to maintaining network stability. Although our model is particularly useful for understanding the dynamic interactions within OSS networks because of the dedicated roles of network members, discussion protocols, and sense-making activity inherent to the act of programming (Brown and Duguid 2002), the results of our study can also be applied to other types of online communities in which participants communicate primarily through online discussion boards. Our main contribution in this study is, therefore, the provision of a theoretical vantage point from which membership dynamics and network stability in the open source community can be better understood.
2. Research Background
We argue that one of the key factors affecting the longevity of an OSS community is the dynamic interaction of its members. The constructual perspective (Carley 1991) suggests that a group's stability is highly dependent on the extent to which group members are structurally able to interact with each other in the pursuit of knowledge sharing and acquisition. Similarly, a resource-based perspective of social structure posits that the social structure must continuously provide members with benefits to ensure its sustainability. Butler (2001, p. 347) argues that "to be sustainable, social structures must maintain access to a pool of resources and support the social process that converts those resources into valued benefits for the participants." The benefits necessary for sustaining the social structure include opportunities for affiliation or championship (McClelland 1985) and opportunities to influence people (Winter 1973), both of which denote the importance of social interactions with affiliated peers. Changes in affiliations or ties (e.g., discontinued membership by an active member) are therefore likely to influence the level of benefits perceived by members, and hence their membership status.
The literature on social capital (Coleman 1988, Burt 1992, Uzzi 1997, Putnam 2000) also stresses the importance of the collective production and reproduction of relationship-based capital. Given that social capital is generated mainly from active interaction and dynamic reciprocity among the members of a community, a lack of participation by or departure of regular members may create a hole in the social fabric by means of which knowledge has been accumulated over time. In this respect, an OSS participant's discontinued participation may eliminate the knowledge conduit that bridges participants with different roles in the development of the project.
The critical role played by such reciprocities in the generation of social capital is reflected in our survey of 34 OSS developers. The majority of the respondents (76%) stated that member dynamics are crucial for the continuous development of OSS projects:
There is much in Gaim [an active OSS community] that I simply cannot do, and without other developers to assist with that, to do those portions of the work, my own work is both fruitless and functionless. (OSS Developer 1) When others are excited about a project and actively contributing, there is a synergy and increased productivity that comes from multiple people working to make a project better. (OSS Developer 2) If key members stop participating, that can affect my total effort level. (OSS Developer 3) It inspires me when there is lots of activity on a project. It gets the creative juices flowing and [helps me] think of new ideas, some of which I am compelled to implement. If a project gets "stale," however, I will usually leave for a similar project given no other compelling reason to stay. (OSS Developer 4)
It should be noted that due to a small sample size, these survey responses may not be generalizable. In addition, factors such as the type of projects and the culture of the community might have influenced the survey responses. Nevertheless, these qualitative data provide some support for the significance of dynamic member interaction for the growth and sustainability of OSS communities. The full details of this survey are available upon request from the first author of this paper.
[FIGURE 1 OMITTED]
Finally, Putnam's stimulating text, Bowling Alone (Putnam 2000), provides a conceptual lens through which the rise and fall of OSS communities could be assessed more thoroughly in conjunction with membership dynamics. This influential text examines the declining trends of civic engagement in America, as reflected in diminishing volunteerism and altruism. Given that OSS communities are also established and maintained by the spirit of volunteerism, the patterns through which traditional volunteer-based associations have evolved and declined may provide a fruitful venue for understanding the life cycle of OSS communities. Putnam's text shows that the rise and fall of membership activities tend to occur rapidly and abruptly, taking the form of an inverted U-shape curve; sudden increases or decreases in terms of membership rates were often observed in civic associations. Interestingly, such abrupt transitions in membership states were also found in several OSS communities in our study (see Figure 1). As demonstrated in Figure 1, after the initial growth period, the participation levels of all four OSS communities (FIPA-OS, Utah-GLX, Bayonne, Galeon) became steady. (4) At a certain point in time, however, all these communities suffered a sudden drop in the number of participants. Although these "phase transitions" (i.e., sudden changes in membership rates) may not be observed in all OSS communities, they nevertheless provide food for thought regarding the role that membership dynamics play in sustaining the communities. In this study, we seek to provide, using the Ising model, a theoretical explanation of such sudden changes in behavioral patterns.
2.1. Ising Model and Herd Behavior
Our goal in this study is to understand, from a quantitative perspective, the basic pattern of interaction among network constituents. We pursue this goal by employing the Ising theory, by which magnetic interactions among microscopic magnets (called "the spin") have traditionally been modeled. These tiny magnets interact magnetically in such a way that their general tendency is to point in a common direction that has a natural mapping to a "herded" state. This tendency is destroyed if outside random noises ("temperature") become high--a state that naturally maps to a chaotic and disorganized state. We will explain these mappings and the concept of temperature in more detail shortly.
Before illustrating the theoretical underpinnings of our study, it is necessary to explain our motives for applying basic principles of physics to the domain of human interaction. Researchers have often argued that there is an operational similarity between physical and social systems (Gunn et al. 1968). Some social phenomena (e.g., organizational survivability, stock market dynamics, employ motivation) have in fact been explained by natural science theories, such as biological evolution and electromagnetism. Many areas of social science--sociology in particular--deal with dynamic social interactions and use qualitative approaches to observe and explicate the complexity inherent to the behavior of social entities. Building on the Ising theory, our study seeks to develop a quantitative model that explains membership dynamics in the OSS community.
The essence of the Ising model is simple and straightforward, yet its conceptual underpinning has been applied to many different contexts, one of which is the exploration of dynamic interactions among market participants. A crucial premise in the application of this theory is that human agents (investors, consumers, and sellers alike) possess an intrinsic desire for consensus (Banerjee 1992, Bikhchandani et al. 1998). The microstructure field extensively employs the principles of the Ising model to illustrate the way in which stock prices fluctuate given various market idiosyncrasies. Herd behavior theory posits that "everyone will be doing what everyone else is doing" (Banerjee 1992, p. 798), leading to the proposition that price movements are determined by the liquid interactions among market participants rather than the arrival of new information (Scharfstein and Stein 1990, Graham 1999). The herding literature suggests that investors tend to engage in nonrational herd behavior, and are driven by an "animal spirit" that has them acting as "imitative lemmings" (Shleifer and Summers 1990, Devenow and Welch 1996).
In his...
|
