Applying Automated Text Analysis to Detect Ideologically-Motivated Manipulations in Mass-Media News

Partisan interest groups manipulate news to cultivate certain viewpoints; the partisan-oriented public creates demand for such polarized information; this forms a harmful feedback loop that amplifies societal polarization. This study presents the application of automated text analysis methodology to detect stylistic and semantic differences between mass-media news sources of different ideological leanings and radicalism levels to identify text characteristics that might indicate information manipulation. 640 articles on four contentious topics from four mass-media news sources, representing combinations of ideological leanings and radicalism levels, were collected from a seven-year period at times of low and high public interest and analyzed using DICTION software. Results showed significant effects of ideological leaning, radicalism level, and the level of public interest on twelve stylistic and semantic text characteristics. These findings were interpreted in the context of possible information manipulations. As a result, we proposed key text characteristics that might indicate information misrepresentation