What do Hamas, Kanye West, and the COVID-19 debate have in common? Well, they seem to provoke anti-Semitism.
Conflicts between Palestinian terrorist organizations and Israel, pandemics, vaccinations, lockdowns, and anti-Semitic rhetoric by celebrities are the main causes of the increase in anti-Semitic messages on Twitter and other social media platforms. is part of
The pandemic has helped unleash an unbridled wave of anti-Semitism online, with Jews blamed not just for COVID-19 but for immigration, culture wars and even racism, settler colonialism and imperialism. As a person, he was criticized by people on the opposite political spectrum.
All this can be observed in real time on social media. The challenge, however, lies in comprehensively monitoring millions of messages across different platforms and languages. Moreover, when stereotyped comments about Jews evolve to become more harmful, pushing Jews out of online or offline spaces or posing physical threats, the impact is even more difficult to determine. . In other words, when does an anti-Semitic message become a threat?
The exact number of anti-Semitic messages sent across the internet every day is unknown, but if we focus only on Twitter and only on English conversations where the word “Jewish” is explicitly mentioned, we estimate that by 2020 was estimated to have over 4,000 anti-Semitic tweets per day.
This number increased in 2021, both in absolute numbers and in the proportion of anti-Semitic messages in conversations about Jews. When people talk about Jews on Twitter, we find that approximately 6 to 20 percent of such conversations are anti-Semitic, depending on the time of day. The highest peak was observed in May 2021 during the violent conflict between Israel and Hamas and Islamic Jihad in Palestine. Fortunately, there are many Twitter users who speak out against anti-Semitism and denounce it in their messages.
Our estimates are based on a representative sample of live tweets categorized using the International Holocaust Remembrance Alliance (IHRA) pragmatic definition of anti-Semitism. Manual classification of samples is an important step towards semi-automated surveillance of online anti-Semitism.
Automatic detection has advanced significantly in recent years with the development of deep learning techniques. The Network Contagion Institute and Strategic Dialogue Institute used some of this technology to detect anti-Semitic messages during Elon Musk’s takeover of Twitter. However, automatically detecting anti-Semitism remains difficult for several reasons.
First, access to data is limited due to platform limitations and the computational power required to process such large amounts of data. Second, the dataset used to train the model is relatively small and does not cover all variations in manifestations of anti-Semitism in the rapidly changing online environment. Third, classification is not always accurate, especially if the annotator does not manipulate live data to understand the message in its “natural” context, including threads, images, links, videos, etc.
Identifying antisemitism is difficult for human annotators, and even harder for machine learning programs at this stage.
We are working on an annotation project to classify messages as anti-Semitic or non-Semitic, and publish data for training machine learning programs. A similar project is also underway, and we will eventually have a large and diverse dataset that can detect anti-Semitic messages with high probability, and monitor their deployment within specific conversations, especially radicalization. is expected to enable We have observed that the conversation can radicalize rapidly in the absence of opposition to anti-Semitic ideas. When norms are established within online communities that view Jews as evil, it’s only a small step for someone to physically attack a Jew.
Anti-Semitism attracts anti-Semites who feel encouraged by it. We need to monitor this cycle, understand its dynamics, and ultimately break it.
