麻豆淫院

CU Boulder political scientist says study underscores the importance of analyzing big datasets to test conventional wisdom

Like many Americans, Alexandra Siegel found herself in a state of shock following Donald Trump鈥檚 victory in the 2016 presidential election. But as a researcher who has worked with vast data sets to explore the digital dimensions of political conflict, she also saw an opportunity.

鈥淚 wanted to do some work that would help us understand the forces that led to and the consequences of the 2016 election,鈥� says , assistant professor of political science at the University of Colorado Boulder.听

By the time she got to CU Boulder, she鈥檇 already built an impressive resume doing just that.

While studying international relations and Arabic at Tufts University, she lived in Jordan as an exchange student. After graduating summa cum laude, with highest thesis honors, Siegel spent a year living in Cairo, Egypt, during the 鈥淎rab Spring,鈥� in which tens of thousands of citizens who demonstrated for democracy made unprecedented use of social-media platforms to coordinate and organize protests.

鈥淚 began to see so much interesting political discourse in online platforms, I wanted to understand what social media can teach us about politics,鈥� says Siegel, who is also a nonresident fellow at the Brookings Institution in the and the , and a faculty affiliate at Stanford University鈥檚 .听

After returning to the United States, she gained experience working in qualitative social media analysis. With an eye toward the quantitative side, she applied to New York University to pursue a PhD and join the school鈥檚 brand-new , where she is now a faculty affiliate.听

鈥淭he lab uses computational social science tools and social media to better understand political dynamics, which wasn鈥檛 really in the popular discourse at the time. They said, 鈥榃e can teach you to do the big-data version of what you鈥檝e been doing,鈥欌�� she says.听

In the wake of Trump鈥檚 election, Siegel zeroed in on Twitter, the former president鈥檚 preferred mode of communication.

鈥淭o what extent did online hate speech and white nationalist rhetoric on Twitter increase over the course of Donald Trump鈥檚 2016 presidential election campaign and its immediate aftermath? The prevailing narrative suggests that Trump鈥檚 political rise鈥攁nd his unexpected victory鈥攍ent legitimacy to and popularized bigoted rhetoric that was once relegated to the dark corners of the Internet,鈥� Siegel wrote as lead author of 鈥淭rumping Hate on Twitter? Online Hate Speech in the 2016 U.S. Election Campaign and its Aftermath,鈥� in the Quarterly Journal of Political Science.听

In search of the answer, Siegal and seven colleagues from New York University and Southern California University analyzed some 1.2 billion tweets鈥�750 million election-related tweets and nearly 400 million tweets from a random sample of American Twitter users.听

鈥�(W)e observe no persistent increase in hate speech or white nationalist language either over the course of the campaign or in the six months following Trump鈥檚 election,鈥� the researchers concluded.听

鈥淲hile key campaign events and policy announcements produced brief spikes in hateful language, these bursts quickly dissipated. Overall, we 铿乶d no empirical support for the proposition that Trump鈥檚 divisive campaign or election increased hate speech on Twitter.鈥�

鈥淚t was,鈥� Siegel says, 鈥渁 very surprising result.鈥�

So surprising, in fact, that the paper took almost four years to be published after being 鈥渄esk rejected鈥濃�攕ent back without being sent out for further review鈥攂y several publications.

鈥淚t鈥檚 hard to know exactly the reason it was so hard to publish,鈥� Siegel says. 鈥淏ut it may be, in part, that the findings weren鈥檛 seen as palatable.鈥�

But she is quick to note that the research was restricted in time and the researchers did not conclude that hate speech didn鈥檛 increase overall:听

鈥淲e just didn鈥檛 see more hateful tweets or more unique users producing that kind of language.鈥� The researchers acknowledge that more extreme users may have gravitated to other platforms at a time when popular platforms were cracking down on hate speech.

Siegel says the findings show the importance of rigorous empirical analysis of large data sets in the social media age.

鈥淏y highlighting the shortcomings of the conventional wisdom regarding the rise of hate speech on Twitter over this period, our paper demonstrates the importance of moving beyond short term or small-scale data sets when studying online speech,鈥� the researchers conclude.听

鈥�(W)e hope future research will explore the extent to which banning accounts may have pushed hateful language o铿� of mainstream platforms and onto more specialized ones, as well as how everyday Internet users encounter and interact with this content on other platforms.鈥�

They also note that, 鈥渙ur analysis of Twitter data tells us nothing about trends in hate crimes, bias incidents, or other o铿刬ne events that have also contributed to the popular narrative of a 鈥楾rump e铿�ect鈥� and deserve further study.鈥�

However, the relative lack of regulation of social media platforms means that researchers are at the mercy of companies when it comes to obtaining data.

鈥淪o much is a black box,鈥� Siegel says. 鈥淚n theory, it would probably be good if external researchers not funded by the platforms could be auditing this kind of thing and putting out peer-reviewed studies rather than the information coming from the platforms.鈥�

Twitter has recently taken some steps to give researchers more access to data, she says, but the world鈥檚 largest platform, Facebook, is 鈥渟till relatively restricted.鈥�

Siegel bucks the recent trend of increasing criticism of social media, arguing that it is a tool with the potential for both positive and negative.听

鈥淚 tell my undergraduate class, 鈥楢uthoritarianism in the Digital Age,鈥� that these tools can be used as positive agents for democratization that give voice to people and at the same time used to exert further control by actors already in power,鈥� she says.听