Media framing is known to be politically biased, but media bias in visual framing under authoritarianism is understudied. This thesis investigates whether Chinese nationalistic media, specifically The Observer (观察者网), differentially selects and frames Trump’s images when compared to the Western media. To examine this, I collected 24,809 images about Trump posted by The Observer and matched them with 232,247 images on Western media through the Google image search. For these paired images, I tailored complementary visual analytics methods including deep learning to calculate image similarities and analyze the differences in visual characteristics. The results show that about 94% of Trump images posted by The Observer are different from those from Western media. Compared to Western media,...[ Read more ]

Media framing is known to be politically biased, but media bias in visual framing under authoritarianism is understudied. This thesis investigates whether Chinese nationalistic media, specifically The Observer (观察者网), differentially selects and frames Trump’s images when compared to the Western media. To examine this, I collected 24,809 images about Trump posted by The Observer and matched them with 232,247 images on Western media through the Google image search. For these paired images, I tailored complementary visual analytics methods including deep learning to calculate image similarities and analyze the differences in visual characteristics. The results show that about 94% of Trump images posted by The Observer are different from those from Western media. Compared to Western media, The Observer tends to post Trump images with higher brightness, higher contrast, and higher visual complexity, while it also depicts Trump as more negative, less likely to laugh, less masculine, and less senior. As a commercialized media, these characteristics in The Observer’s visual framing may promote the information spread and realize its commercial interest. As a propaganda media, these visual characteristics depict Trump as an incapable and negative politician, catering to its nationalistic audience. This thesis contributes to the literature on media visual framing, media bias under authoritarianism, and methodologically to visual analysis in computational social science.

Keywords: visual framing, media bias, automated visual analytics, computational social science