Online Library

Hate speech detection on Twitter is critical for applications like controversial event extraction, building AI chatterbots, content recommendation, and sentiment analysis. We define this task as being able to classify a tweet as racist, sexist or neither. The complexity of the natural language constructs makes this task very challenging. We perform extensive experiments with multiple deep learning architectures to learn semantic word embeddings to handle this complexity. Our experiments on a benchmark dataset of 16K annotated tweets show that such deep learning methods outperform state-of-the-art char/word n-gram methods by ~18 F1 points.

VOX-Pol Blog post.

A balkanized Internet beset by hostile cyber-related activities raises a host of questions and problems for the U.S. government, American corporations, and American citizens. The Council on Foreign Relations launched this Task Force to define the scope of this rapidly developing issue and to help shape the norms, rules, and laws that should govern the Internet. This is the report published by the Task Force.

Following Sahana Udupa and Matti Pohjonen’s (2019) invitation to move the debate beyond a normative understanding of hate speech, this article seeks to build a foundation for conceptual and empirical inquiry of speech commonly considered deviant and disturbing. It develops in three stages. It first maps the public lives of terms that refer to online vitriol and how they have been used by different communities of researchers, politicians, advocacy groups, and national organizations. Second, it shows how different types of “haters” have been interpreted as parts of “swarms” or “armies,” depending on whether their violent potential emerges around critical incidents or whether they respond to longer-term strategies through which communities and their leaders tie their speech acts to explicit narratives. The article concludes by locating “extreme speech” within this broader conceptual tapestry, arguing that the paternalistic the gaze that characterizes a lot of research on online hate speech is tied to what Chantal Mouffe has referred to as the “moralization of politics,” a phenomenon that cannot be matched by responses that are themselves moral.

Amazon has been called the “everything store,” but today it is much more than just a store, with publishing, streaming, and web services businesses. Its reach and infuence are unparalleled: Most U.S. online shopping trips begin at Amazon, Amazon dominates the U.S. e-book business, and the company’s web services division has over 60 percent of the cloud computing services market. All this adds up for Amazon and its owners. The company posted record profts of $1.9 billion in the last three quarters of 2017,7 and CEO Jef Bezos’s wealth soared to $140 billion in 2018, largely because of the value of Amazon stock. A close examination of Amazon’s various platforms and services reveals that for growing racist, Islamophobic, and anti-Semitic movements, the breadth of Amazon’s business combined with its weak and inadequately enforced policies provides a number of channels through which hate groups can generate revenue, propagate their ideas, and grow their movements. We looked at several areas of Amazon’s business, including its online shops, digital music platform, Kindle and CreateSpace publishing platforms, and web services business.

Extreme, anti-establishment actors are being characterized increasingly as ‘dangerous individuals’ by the social media platforms that once aided in making them into ‘Internet celebrities’. These individuals (and sometimes groups) are being ‘deplatformed’ by the leading social media companies such as Facebook, Instagram, Twitter and YouTube for such offences as ‘organised hate’. Deplatforming has prompted debate about ‘liberal big tech’ silencing free speech and taking on the role of editors, but also about the questions of whether it is effective and for whom. The research reported here follows certain of these Internet celebrities to Telegram as well as to a larger alternative social media ecology. It enquires empirically into some of the arguments made concerning whether deplatforming ‘works’ and how the deplatformed use Telegram. It discusses the effects of deplatforming for extreme Internet celebrities, alternative and mainstream social media platforms and the Internet at large. It also touches upon how social media companies’ deplatforming is affecting critical social media research, both into the substance of extreme speech as well as its audiences on mainstream as well as alternative platforms.