Thread by @ErikBleich1

Our book, a 🧵.

1/ We hope Covering Muslims will engage scholars, students, activists, journalists & anyone interested in portrayals of Muslims, media analysis, or big data.
📢 Please RT.

If this 📖 has info you want/need, we want you to have a copy. Read on!

@amauritsvdv

2/ Our book has several key findings. But most importantly, there is one that can’t be overstated:

👉 COVERAGE OF MUSLIMS IS REMARKABLY NEGATIVE BY ANY MEASURE 👈

3/ We provide the first systematic, large-scale analysis of American newspaper coverage of Muslims, using big data techniques and readings of individual articles. 📈

4/ We compare coverage of Muslims in the US to coverage of Catholics, Jews, Hindus, African Americans, Latinos, Mormons, and atheists.

➡️ Bottom line: coverage of Muslims is FAR more negative than for any other group.

5/ We analyze articles from 1996 through 2016—looking specifically at the effects of 9/11 and other major events on the number and tone of articles, as well as the words most commonly associated with Muslims.

6/ We compare coverage in the US🇺🇸 to that in Britain🇬🇧, Canada🇨🇦, and Australia🇦🇺, as well as to newspapers in India🇮🇳, Kenya🇰🇪, Malaysia🇲🇾, Nigeria🇳🇬, Pakistan🇵🇰, and Singapore🇸🇬.

7/ We use topic modeling techniques to explore the themes most commonly associated with Muslims, identifying the ones that are the most (and least) negative.

8/ We conclude by thinking through the ways in which media outlets reproduce biases—especially Islamophobia—and how media consumers can “tone-check” ✅ the media to offset some of those biases.

9/ Stay tuned for short threads about each of these main findings in the coming days… 📅

(📢 Retweet please and thank you!)
(❓Questions/comments welcome!)

10/ One main comparison we examine is coverage across groups.

Is coverage of Muslims really more negative than it is of other groups?

YES. By a long shot. One way to see this is by comparing the % negative and positive stories mentioning Muslims, Hindus, Jews, and Catholics.

11/ Similarly, stories touching on Muslim Americans are more likely to be negative than those mentioning different US-based racial, ethnic, and (non-)religious minorities.

12/ We use a new lexical sentiment analysis tool, MultiLexScaled, to measure not only whether an article is positive or negative, but also just how positive or negative it is benchmarked against a random sample of over 48,000 US newspaper articles. 🤓

13/ This allows us to show that articles mentioning Muslims are far more negative than those touching on other groups. (The x-axis is standard deviations.)

14/ Coverage of most other groups is far less negative. Muslims are a striking outlier.

15/ In short, YES, coverage of Muslims is undeniably and strongly more negative than that of other groups.

In fact, the average article about Muslims is more negative than 84% of our random sample of US newspaper articles. No other group comes close.

16/ We also looked at change over time. ⏲️ How big a deal was 9/11 for the newspaper coverage of Muslims?

Surprisingly, not much in terms of overall negativity, as we see if we chart valence from 2000-2003.

17/ However, 9/11 did mark an enormous jump in the prevalence of references to terrorism and extremism.

18/ This is clear, too, if we look at the words most commonly appearing right next to “Muslim” or “Islam” in newspaper articles since 9/11, shown in this word cloud (terrorism and extremism words are highlighted in black).

19/ We then looked at geographic comparisons in coverage.

If Muslim articles in the United States🇺🇸 are so negative, is that unusual? We check by compiling comparable sets of articles from 1996 to 2016 from Britain🇬🇧, Canada🇨🇦, and Australia🇦🇺.

20/ The answer is NO. Coverage of Muslims in these four countries is very similar. And similarly negative compared to our representative sample of articles.

21/ Words like “radical,” “fundamentalist,” “militant,” and “extremist” are among the 10 most common words associated with Muslims or Islam in all four countries.

22/ The picture changes a bit when looking at newspapers in India🇮🇳, Pakistan🇵🇰, Kenya🇰🇪, Nigeria🇳🇬, Singapore🇸🇬 and Malaysia🇲🇾. The Kenyan newspaper is JUST as negative. The Malaysian newspaper is on average POSITIVE. The rest are in between.

23/ World events do not dictate negative coverage with respect to Muslims. But developed Anglophone countries are uniformly extremely negative. Why is that? 🤔

24/ A topic model shows that a large share of Muslim coverage is foreign coverage 🌐 (news events in Muslim-majority countries such as those in the Middle East). Previous scholarship shows that foreign news tends to be negative, and we find this as well.

25/ Not all of this news is negative _about_ Muslims. About 3% of all coverage of Muslims is linked to the former Yugoslavia, Uighurs (China🇨🇳), and Rohingya (Myanmar🇲🇲). These articles tend to portray Muslims as victims of violence and oppression.

26/ Also, not all topics of coverage are extremely negative. Articles about culture and education are close to neutral in valence, on average. This includes the article by @lydiakiesling with the art by @maxomatic that is featured on our book’s cover: nytimes.com/2016/10/09/mag…

27/ Somewhat surprisingly, news articles focused on religious topics are also much less negative than the average article in our dataset.

28/ On the whole, though, 👉articles about Muslims and Islam are strikingly negative— 👉negative compared to other groups, across time, across Anglophone countries, and across topics.

29/ This raises the question of whether journalists are perpetuating Islamophobia. Even if they don’t mean to have that effect, their rhetoric has a cumulative impact. Negativity matters.

30/ We show that negative coverage affects individual attitudes in forthcoming experimental research in @PandRJournal. We argue that readers need to “tone-check” ✅ the media.

31/ When you read a story about Muslims, what words or allusions trigger a negative reaction?

There are A LOT of these for Muslim articles. And for other marginalized groups.

You can partly offset that instinct by realizing just how negative the story is. 🧐

32/ For those interested in the methods, the sentiment analysis tool we used to assess the tone of Muslim coverage, MultiLexScaled, is available at GitHub: github.com/amaurits/Multi….

33/ It includes python code and notebooks, along with the scaling parameters used to calibrate and benchmark valence measures.

34/ The method outperforms individual lexica as well as some machine learning classifiers across a range of validation tests, and can be applied off-the-shelf. The code at GitHub also includes a text cleaning function to preprocess texts. 🧹

35/ Finally, for those who stuck with this to the end, we want you to be able to read the 📖.

💥✨If you are a student or scholar without a research budget or a way to easily access the book, email one of us. We will try to get you a copy. ✨💥

/end