PURPOSE: Previous research has called for media balance when reporting on chronic traumatic encephalopathy (CTE) in order to avoid harmful bias towards readers. This call is raised from concerns that the media’s representation of CTE has moved beyond what science has proven. The purpose of this study was to use word sentiments to directly compare journal articles with corresponding news articles to evaluate these concerns. The news articles were split into three groups: press releases reporting the articles’ findings, news articles about CTE from upper tier news outlets, and articles from lower tier news outlets. METHODS: Research articles (n=10) directly associated with CTE that were heavily covered in the media were selected for this sample. An equivalent number of press releases (n=10), upper tier articles (n=10), and lower tier articles (n=10) were collected in order to compare semantics. The “AFINN” sentiment analysis dictionary rates the emotional valence of each word with an integer between minus three (negative connotation) and plus three (positive connotation). Words not recognized by the dictionary or with a zero weight were omitted from the analyses. Mean sentiment score was adjusted for total word count. RESULTS: The mean sentiment scores, adjusted were words count, were as follows: 0.086 for journal articles, -0.096 for press releases, -0.122 for upper tier sources, and 0.026 for lower tier sources. An analysis of variance calculation yielded no significant differences between the groups (F = 1.058, p = 0.379). CONCLUSIONS: Despite recent calls for a less biased reporting of CTE in mainstream media, our analysis indicates essentially equal sentimental weighting between peer-reviewed journal articles and news reports on CTE, whether the report was a press release, an article from an upper tier source, or from a lower tier source. Additionally, these sentiment weights each approached a value of zero (true neutrality). Future research should take into account the context in which the words appears in the articles in addition to using sentiment averages.
Read full abstract