A tripartite methodology was utilized to evaluate the research question. Initially, there were plans to adapt a model of the Implicit Association Test which has been used to test implicit biases. However, this had a number of shortcomings — primarily, it was not easy to disseminate and I did not have the technical knowledge to adapt this model. Thus, a model built off of surveys were utilized. There were three parts to this survey — an initial survey of views, a reading of an article, and a final survey of views.
I initially wanted to do multiple surveys, with multiple articles being read by each person. However, my research advisor mentioned that increasing statistical significance would be easier with one survey — it would allow for increased dissemination. Thus, I rethought my stance on having multiple articles being given out. Instead, I chose to do one article that argued against the stereotypes that were the most present in the media.
In order to determine which stereotypes were the most present, I initially had plans to conduct a corpus based computational analysis of media, and generate stereotypes in the media based off of that. However, there was already a study which had done that. In order to decrease inherent biases, this research was grounded in past work that had actually conducted this corpus based linguistics analysis. Specifically, the work that determined these stereotypes was Negative Image Construction of North Korea: Nuclear Orientalism in the U.S. Newspapers by Binnarae Oh. Writing for the University of Minnesota, uses a corpus based collocative study of US newspapers from the past five year, and analyzed common words and themes that were present in these articles. Due to this objective stance, as he utilized a computer based method of sorting rather than sorting the papers by hand or through individual determinations of rhetoric, I found it to be the most replicable. This also makes this paper more replicable as another person who holds computational skills can perform the same action and generate the same stereotypes. From Oh’s paper, the stereotypes that were the most common were: instability of the country and insanity or craziness of the regime and its leaders. These were done by grouping the words and phrases which featured most heavily in Oh’s paper with similar meaning together.
Then, I chose an article from Foreign Policy that argued against these stereotypes. It predominantly argued against calling North Korea crazy. This paper was selected for the survey not only because it argued against the primary stereotypes that were mentioned in Oh’s analysis, but as it had a very methodical and logical reasoning compared to other articles that argued a similar point and had the potential to convince readers.
The survey was constructed prior to selecting the article so as to avoid internal biases. It was a fully anonymous survey. The first portion of the survey was a set of demographic questions and questions designed to test initial views about North Korea. It also asked questions about how often respondents read the news, the time they spent reading the news, whether or not they utilized social media for news, and which news sites they gained their information from. After this survey, readers read the Foreign Policy article previously mentioned. The article had all identifying information about the author and the news website cut out — only the text was presented. Although this may detract from the article, as respondents may not believe it without identification information, I chose to do this as it provided the advantage of removing internal biases that respondents may have held about the source or author of the article. After reading the article, the respondents were given a final survey. This focused on asking them how their views changed.
During an initial pilot survey with six participants, I found that most of my respondents views had no change. Initially, I believed that people’s views wouldn’t be changed by simply reading an article, and that I would have no results to showcase. However, after careful reflection, I decided to ask further questions instead of a simple yes or no. I chose to ask which views changed, and which views did not change. Although these responses were more open ended and were harder to analyze quantitatively, it provided more significant information to analyze.
Google Surveys was utilized to perform the survey, and Google Spreadsheets was utilized to compile and process the data. I processed the raw data through a variety of methods. For the multiple choice questions, analyzing them quantitatively was simple as it was presorted. However, for the qualitative data, I utilized a variety of methods to process the data. I used dictionary definitions to group the word characterizations together, and once I had them grouped, I generated a theme that encompassed the characterization. These were then counted to show how much these themes recurred.