Communicative exchanges consist of a certain degree of both static and dynamic structure that can be used for prediction. Temporal dynamics are often neglected in communication studies. We use Shannon’s mathematical theory of communication to examine 1,594,690 contributions from 206,184 contributors to 38 open online collaborations. We find that about three fourths of the total predictability of turn taking stem from participation frequencies (‘static variance’), while one fourth originates from the temporal sequence (‘dynamic process’). Most dynamic structure is contained within consecutive dyads. We find a trade-off in the importance of static and dynamic structure, which we explain with a combination of both theoretical and empirical factors. We also show that the stationarity of the communication process plays a significant role in this trade-off. These findings have implications both for theorizing and methodologically measuring communication as a dynamic process, as well as for the practical design of online collaboration systems. Acknowledgements: We are indebted to Jim Crutchfield for the continuous exposition to the depth and beauty of information theory and its extensions, as well as to Robert Bell, Peter Monge, Sarah Marzen, Grace Benefield, Harrison Hughes, and reviewers of IC2S2 for helpful comments. STATIC STRUCTURE AND TEMPORAL DYNAMICS 2 The Complementary Importance of Static Structure and Temporal Dynamics in Teamwork Communication Communication patterns of turn taking of team communication are complex, adaptive, and emergent phenomena. The basic signature of complex adaptive systems is their situating between structure and randomness: some aspects of them are predictable, while others are inherently random (Bialek et al., 2001; Crutchfield and Feldman, 2003; Crutchfield, 1994; Grassberger, 1986; Kolmogorov, 1959). We use an approach from the literature of complex systems, which has grown out of Shannon’s (1948) mathematical theory of communication, to quantify the amount and kind of predictability in turn taking of teamwork communication. The goal is to deepen our understanding of its origins, constituents, and involved trade-offs. Static Structure and Dynamic Patterns The literature differentiates between varianceand process-based explanations of communication patterns (Barnett, Chang, Fink, and Richards, 1991; Monge et al., 1984; Poole, 2007; Poole et al., 2000). The former emphasizes relations among different variables during a given time window and the latter among variables at different points in time. Today, the vast majority of the literature analyzes variables or networks derived from a static snapshots or a time-collapsed sequence of communication processes (Keegan, Lev, & Arazy, 2016; Leenders, Contractor, & DeChurch, 2016; Monge et al., 1984; Pilny, Schecter, Poole, & Contractor, 2016). This answers the ‘who speaks how often’ question, but neglects the ‘when’ question. For example, we can calculate the share of contributions of users in Wikipedia and find that 10% of users make more than 90% of the contributions (Ortega, Gonzalez-Barahona, and Robles, 2008). This provides predictability: it is quite likely that the next edit will come from a top power-user. In this case, the entire editing history is seen as one static event. The process-based approach recognizes that “communication is a process and should be explained as such” (Poole, 2007, p. 181). For example, we might notice that the contribution of a sporadic ad-hoc user might make it more or less likely that a power-user gets involved. Here we use dynamical patterns for predictions, conditioning on temporal sequences. The predictability of the next communicative turn depends on the particular historical context that immediately precedes and currently frames the present exchange. In the early 2000, less than 10% of articles in 40 communication journals dealt explicitly with temporal dynamics (Poole, 2007). The current neglect of temporal dynamics in the communication literature exists despite a focus on processes in early works (Berlo, 1960; Schramm, 1955), and despite a big push in the 1970s and 1980s to use dynamical systems theory, which included the study of group decision-making (Ellis and Fisher, 1975; Fisher, 1970; Krain, 1973; Poole and Roth, 1989), relational control in relationships (Ellis, 1979; Fisher and Drecksel, 1983; Hawes and Foley, 1973), mass communication (Watt and VanLear, 1996), and talk and silence sequences in conversations (Cappella, 1979, 1980; Cappella and Planalp, 1981). A main STATIC STRUCTURE AND TEMPORAL DYNAMICS 3 historical account for why this work was discontinued was that “gathering everyday conversations... is nearly impossible.... unless one carries a tape recorder around all day (a cumbersome and hardly practical endeavor)” (Fisher and Drecksel, 1983, p. 68). Additional culprits are the “adoption of approaches from other fields such as psychology that do not emphasize process as much as communication” (Poole, 2007, p. 181), “the perceived scope of effort required from the researcher” (Monge et al., 1984, p. 28), and that dynamics were “simply impractical to compute” (Attneave, 1959, p. 22) before today’s computing power.

中文翻译：

---

团队沟通中静态结构和时间动态的互补重要性

交际性交流包括可以用于预测的一定程度的静态和动态结构。在交流研究中，时间动态常常被忽略。我们使用Shannon的交流数学理论，检查了206,184位贡献者对38个在线开放协作的1,594,690位贡献。我们发现，转弯总可预测性的大约四分之三来自参与频率（“静态方差”），而四分之一则来自时间序列（“动态过程”）。大多数动态结构包含在连续的二元组中。我们在静态和动态结构的重要性之间进行权衡，并结合理论和经验因素对此进行了解释。我们还表明，沟通过程的平稳性在这种权衡中起着重要作用。这些发现对于理论化和方法论上将交流作为一个动态过程进行测量以及对在线协作系统的实际设计都具有启示意义。致谢：我们谨向吉姆·克劳奇菲尔德（Jim Crutchfield）不断阐述信息理论及其扩展的深度和美感，以及罗伯特·贝尔，彼得·蒙格，莎拉·马岑，格蕾丝·贝内菲尔德，哈里森·休斯以及IC2S2的审稿人。静态结构和时间动态2团队沟通中静态结构和时间动态的互补重要性团队沟通的轮流沟通模式是复杂的，适应性的和突发的现象。复杂的自适应系统的基本特征是它们位于结构和随机性之间：它们的某些方面是可预测的，而其他方面固有地是随机的（Bialek等，2001； Crutchfield和Feldman，2003； Crutchfield，1994； Grassberger，1986； Kolmogorov （1959年）。我们使用来自Shannon（1948）的通信数学理论的复杂系统文献中的一种方法来量化数量和可预测性，进而进行团队协作通信。目的是加深我们对它的起源，组成和所涉及的折衷的理解。静态结构和动态模式文献对方差和基于过程的通信模式解释进行了区分（Barnett，Chang，Fink和Richards，1991； Monge等，1984； Poole，2007； Poole等，2000）。前者强调给定时间窗口内不同变量之间的关系，而后者则强调不同时间点处的变量之​​间的关系。如今，绝大多数文献都分析了从静态快照或通信过程的时间崩溃序列中衍生的变量或网络（Keegan，Lev和Arazy，2016； Leenders，Contractor和DeChurch，2016； Monge等， 1984年； Pilny，Schecter，Poole和Contractor，2016年）。这回答了“谁说多久一次”的问题，却忽略了“何时”问题。例如，我们可以在Wikipedia中计算用户贡献的份额，发现10％的用户贡献了90％以上的贡献（Ortega，Gonzalez-Barahona和Robles，2008年）。这提供了可预测性：下一次编辑很有可能来自高级用户。在这种情况下，整个编辑历史记录被视为一个静态事件。基于过程的方法认识到“沟通是一个过程，应该这样解释”（Poole，2007，第181页）。例如，我们可能会注意到，零星的临时用户的贡献可能会或多或少地使高级用户参与其中。在这里，我们使用动态模式进行预测，并以时间序列为条件。下一轮交往的可预测性取决于紧随当前交流并当前构成当前交流的特定历史背景。在2000年初，在40种通讯期刊中，只有不到10％的文章明确涉及时间动态（Poole，2007年）。尽管关注早期作品中的过程，但目前对传播文学中时间动力的忽视仍然存在（Berlo，1960； Schramm，1955），尽管在1970年代和1980年代大力推动使用动力学系统理论，其中包括对群体决策的研究（Ellis和Fisher，1975； Fisher，1970； Krain，1973； Poole和Roth，1989），关系（Ellis，1979; Fisher和Drecksel，1983; Hawes和Foley，1973），大众传播（Watt和VanLear，1996），交谈中的谈话和沉默序列（Cappella，1979，1980; Cappella和Planalp，1981）。静态结构和时间动态3的主要历史原因是：“为什么不进行日常对话……几乎不可能……。除非一整天都携带一台磁带录音机（这是一件繁琐且几乎不实际的工作）” （Fisher and Drecksel，1983，第68页）。