As a check on the validity of utilizing the netlit
approach to creating literature reviews, we conducted an experiment in which five (junior) researchers independently coded the same set of articles and then compared the resulting networks. The researchers read fifteen articles, pulled from a section of James Vreeland’s International Organizations syllabus. No researcher had prior knowledge or specialization in international organizations. We expect this exercise to be representative of researchers in a higher variance group given low correlation of the limited prior knowledge on the topic (and therefore a harder test). Each researcher followed the standard netlit
data entry approach and coded the articles’ relationships in a simple Excel spreadsheet containing three columns: from, to, and citations. Using the resulting spreadsheet, we then used the review() function from our netlit package to create the network diagrams for each researcher, which are shown below.
Researcher 1 | Researcher 2 | Researcher 3 | Researcher 4 | Researcher 5 | |
---|---|---|---|---|---|
Researcher 1 | NA | 0.50 | 0.55 | 0.68 | 0.62 |
Researcher 2 | 0.93 | NA | 0.75 | 0.68 | 0.74 |
Researcher 3 | 0.96 | 0.62 | NA | 0.71 | 0.82 |
Researcher 4 | 0.74 | 0.54 | 0.85 | NA | 0.64 |
Researcher 5 | 0.89 | 0.65 | 0.75 | 0.71 | NA |
Researcher 1 | Researcher 2 | Researcher 3 | Researcher 4 | Researcher 5 | |
---|---|---|---|---|---|
Researcher 1 | NA | 0.42 | 0.40 | 0.55 | 0.49 |
Researcher 2 | 0.67 | NA | 0.50 | 0.45 | 0.59 |
Researcher 3 | 0.44 | 0.29 | NA | 0.58 | 0.31 |
Researcher 4 | 0.26 | 0.27 | 0.55 | NA | 0.33 |
Researcher 5 | 0.63 | 0.46 | 0.65 | 0.66 | NA |
These quantitative results are supported by the qualitative feedback we elicited from the researchers. We asked the five researchers “if you were to separately use each of these networks to understand the current state of the literature – both what currently exists and what is missing – would you come away with similar conclusions?” and received positive responses from all five. For example, Researcher 3 stated that “after reading the 15 papers provided and then looking at everyone’s maps, the central causal relationships are present in all maps.” Research 5 reported that “except for some minor details, everyone captured similar causal relationships as the main ones.” This feedback increases our confidence that our method is replicable; starting from the same literature, five independent junior researchers with no prior knowledge on the topic arrived at similar conclusions regarding the state of the literature. In addition, these represented the modal type of researcher who might use our review method: junior researchers who are familiar with the discipline but are new to a specific topic.