Viewpoints regarding the “bogus information”
To answer you to definitely matter, we again reviewed the new responses subjects provided when expected exactly what phony information and propaganda indicate. I analyzed solely those responses in which subjects considering a meaning to possess either name (55%, letter = 162). Note that the fresh proportion out of subjects whom considering particularly meanings is actually lower than inside the Studies 1 (95%) and you will 2 (88%). Upon closer examination, we discovered that multiple subjects had most likely pasted definitions away from a keen Google search. Into the an exploratory analysis, we located a mathematically significant difference regarding the likelihood one to participants provided a great pasted definition, centered on Political Personality, ? 2 (dos, Letter = 162) = seven.66, p = 0.022. Specifically, conservatives (23%) were likely to be than simply centrists (6%) to include an effective pasted meaning, ? dos (step 1, N = 138) = seven.31, p = 0.007, Otherwise = cuatro.57, 95% CI [step one.29, ], every other p beliefs > 0.256. Liberals fell ranging from these extremes, that have thirteen% getting a pasted meaning. As we were trying to find subjects’ own meanings, i omitted this type of doubtful responses away from study (letter = 27).
We followed a similar analytic procedure as with Tests step one and you can dos. Desk cuatro displays such studies. Due to the fact table shows, brand new proportions of victims whose answers included the characteristics demonstrated from inside the Try out 1 was in fact comparable around the political personality. Specifically, we did not simulate the fresh shopping for out-of Experiment step one, which individuals who known leftover had been expected to render separate meanings towards conditions than simply people that understood right, ? dos (step one, Letter = 90) = step one.42, p = 0.233, virtually any p thinking > 0.063.
Additional exploratory analyses
We now turn to our additional exploratory analyses specific to this experiment. First, we examine the extent to which people’s reported familiarity with our news sources varies according to their political identification. Liberals and conservatives iliar with different sources, and we know that familiarity can act as a guide in determining what is true (Alter and Oppenheimer 2009). To examine this idea, we ran a two-way Ailiarity, treating Political Identification as a between-subjects factor with three levels (Left, Center, Right) and News Source as a within-subject 
factor with 42 levels (i.e., Table 1). This analysis showed that the influence of political identification on subjects’ familiarity ratings differed across the sources: F(2, 82) = 2.11, p < 0.001, ? 2 = 0.01. Closer inspection revealed that conservatives reported higher familiarity than liberals for most news sources, with centrists falling in-between (Fs range 6.62-, MRight-Leftover range 0.62-1.39, all p values < 0.002). The exceptions-that is, where familiarity ratings were not meaningfully different across political identification-were the media giants: The BBC, CNN, Fox News, Google News, The Guardian, The New York Post, The New York Times, The Wall Street Journal, The Washington Post, Yahoo News, and CBS News.
We also predicted that familiarity with our news sources would be positively associated with real news ratings and negatively associated with fake news ratings. To test this idea, we calculated-for each news source-correlations between familiarity and real news ratings, and familiarity and fake news ratings. In line with our prediction, we found that familiarity was positively associated with real news ratings across all news sources: maximum rGenuine(292) = 0.48, 95% CI [0.39, 0.57]; minimum rReal(292) = 0.15, 95% CI [0.04, 0.26]. But in contrast with what we predicted, we found that familiarity was also positively associated with fake news ratings, for two out of every three news sources: maximum rBogus(292) = 0.34, 95% CI [0.23, 0.44]; minimum rFake(292) = 0.12, 95% CI [0.01, 0.23]. Only one of the remaining 14 sources-CNN-was negatively correlated, rFake(292) = -0.15, 95% CI [-0.26, -0.03]; all other CIs crossed zero. Taken together, these exploratory results, while tentative, might suggest that familiarity with a news source leads to a bias in which people agree with any claim about that source.