... science1.1
The classification is not meant to be exclusive. A paper can have characteristics of more than one approach at the same time. I also read and present the papers so that they exemplify patterns relevant for an approach.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Science1.2
http://symposium.computationalsocialscience.eu/2019/
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... solved.2.1
Algorithms, as used in this book, should not be confused with the various meanings given to algorithms in critical algorithm studies, which may be more familiar to social science students. Critical algorithm studies highlight the importance of algorithms as part of the digitalised and software-enabled society (e.g. Kitchin, 2017; Gillespie, 2012; Bucher, 2012). As Kitchin (2017) summarises well, this research area seeks to explore how humans create and maintain this society and often study code or algorithms as part of this process. However, as the work by Seaver (2017) highlights, social scientists focus on wider socio-technical systems that are labelled as algorithms in the research literature. Following his call, when referring to critical studies on algorithms, we will call them algorithmic systems and not algorithms.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... R2.2
This is to say, if you seek to learn how to conduct statistical analysis with R, this book does not cover those details. Instead, I recommend looking at books such as xxx. I have tried to show some of the shortcuts available at R in the code examples, but the procedural paradigm does not always fit well with it.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Data.3.1
The model has extremely poor accuracy. Therefore, the model should be seen as an example only.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... formats4.1
Several network analysis software have unique data formats used for storing networks. For example, Gephi (https://gephi.org/) uses GEXF while Ucinet outputs UCINET DL. We focus on these data formats because they are non-proprietary. Many network analysis tools know how to import or export data from these formats.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... springs.4.2
As these algorithms are complex, we do not show example codes for network layout algorithms or many other more difficult algorithms discussed throughout this book. Writing these complex algorithms is demanding and difficult but also time consuming. Therefore, researchers often use code from other people (libraries; see Section 8.2) instead of their own code.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... system7.1
Alternatively, we can create a `bubble' of the future where we study novel technologies through artificially injecting and sustaining them (Salovaara et al., 2017). This requires running a field deployment where the novel technology is available, which is not that different from the field experiments we discussed above. However, to allow this to happen, the prototype must be functional and act as a research product instead of a prototype (Odom et al., 2016).
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... posts.9.1
The observed Cohen's effect sizes were rather small, $d < 0.05$. Commonly, the effect at this threshold level would be clearly small ($d=0.20$). This suggests that while the study was able to observe a statistically significant difference between control and treatment groups, this is most likely thanks to its massive size with over a half million participants. The validity of the research finding is its own debate, also leading to discussion and dialogue (e.g. Panger, 2016).
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.