How do you fight AI hate? With AI

The latter, particularly, underlaid propaganda disseminated during the recent Israel-Hamas conflict to delegitimise Israel across the international community.

AI-generated content has also been used to discredit Jewish figures and falsify Jewish history.

Antisemitic social media posts are, though, an incessant threat, especially when they gain multiple likes and shares. It is bad enough when they are written by Jew-haters.

But a significant numbers are generated by bots and posted from fake accounts. Before Elon Musk bought Twitter (X) in 2022, he employed anti-disinformation firm Cybara, which estimated that around 11 per cent of user activity was bot-driven. “We will defeat the spam bots or die trying,” said Musk at the time. But, by all accounts, this percentage has increased considerably since.

Whether the source is a real person or a bot, social media platforms have generally failed to remove the proliferation of anti-Jewish posts. It’s led several organisations and individuals to focus on tackling the problem. The rationale is simple: to fight an AI threat you need an AI response.

In 2022, the Anti-Defamation League (ADL) produced the Online Hate Index. It was, it claimed, the first-ever independent AI tool for detecting online antisemitism.

Chief executive Jonathan Greenblatt said: “We will use this tool to hold social media platforms accountable for how well they proactively take down hate and how well their content moderators respond to reports [of it].” In its first runs the tool found that Reddit and Twitter failed to remove more than 70 per cent of antisemitic content. Two months later, despite informing the platforms, more than 56 per cent was still there.

Recognising social media companies’ inertia, some initiatives look to achieve greater platform penetration. A monitoring tool from Fighting Online Antisemitism (FOA) tracks antisemitic content. Once it finds something, FOA, like the ADL, aims to inform the platform. As Yuval Pecht, its project manager, explained, it does so from the position of having “trusted flagger status”, which should allow them to be taken more seriously.

AI tools employed by the ADL, FOA and organisations using similar technology may be good at detection, accelerating that process beyond human capability. But they still require a human for dealing with platforms and their moderators – the stage where barriers are often hit.

Savee takes a different approach. Its AI tool is in the form of an app downloadable onto a user’s computer. Bear in mind most people will be unable to keep writing fresh responses to antisemitic content and with all the needed facts. This tool offers users a AI-generated response to Holocaust denial, anti-Israel propaganda and so forth at the press of a button. The user can then post that response. Savee is good at what it does but its impact can only be proportional to the number of its individual users – and their stamina!

However effective antisemitism detection is, the scale at which these tools operate is limited. And online Jew-hate consequently remains an issue increasing faster than the Jewish community’s response. The tools also frequently rely on volunteers or user goodwill.

The work of MIT students Avi Balsam and Zack Duitz, therefore, represents an important advance. In reaction to the rising anti-Israel narrative following October 7, they developed Project Chazon.

Their creation detects anti-Israel content, generates a response, and automatically posts it back to the platform. To achieve that level of automation, users are asked to create a social media account and effectively hand it over. The tool then takes over completely. Currently, it works only on X, and requires real accounts rather than fake ones to avoid falling foul of the platform’s filters. It’s a prototype but the potential is there, the developers tell me, for it to become multi-platform. Project Chazon has reached the semi-finals of a contest called the Adir Challenge, which asks entrants to find original ways to fight antisemitism.

Extending Project Chazon’s concept, at CAAI (Combat Antisemitism with AI) we’ve shown it is feasible for AI to respond to antisemitic comments, as opposed to only anti-Israel comments. ChatGPT enables anyone to test multiple simulations at virtually no cost; though to take such work beyond prototyping still requires resourcing.

Academics are also working on AI-based initiatives. The largest is the international Decoding Antisemitism Project led by Mathias Becker of the Technical University, Berlin.

One of its aims is to find ways to identify coded, or implicit, antisemitic speech – speech that could be in the form of metaphors, allusions, or abbreviations.

For example, at a basic level, codes beloved by the far right such as “JQ” (Jewish Question), or “Rothschild” when referring to the stereotype of Jewish financial power rather than the family’s name.

Smaller projects exist too. Nathalie Japkowicz at the American University in Washington DC recently began developing an AI tool that can, through online monitoring, determine the latest popular antisemitic slur. And they do change. Take the recent attacks against Israeli football fans in Amsterdam where the previously unheard chant “cancer Jew”’ rang out.

For addressing attempts to falsify Jewish history, AI has been a boon. The technology has been used by organisations such as the Holocaust Educational Trust to develop “survivor testimony platforms”.

Recreated as a simulation, the survivor can tell their story and interact with questioners.

It is a first-person record that will continue beyond the life of the survivor. Generative AI for Good run by Shiran Mlamdovsky Somech uses similar technology to enable Warsaw Ghetto fighters, recreated from old pictures, to tell their stories.

These initiatives are amazing educational tools where, by telling what actually happened to Jews, they reinvigorate the historical narrative. But the extent they – as with other types of Holocaust education – can reduce antisemitism is part of a long-standing debate.

Nevertheless, these are all essential endeavours and must continue. Yet, the toxic atmosphere of the online world now experienced by Jews remains. And it prompts questions. Chiefly, are current AI initiatives sufficient to fight back? And what more could be done?

In reality, with much current AI work focused on anti-Jewish content detection, with the exception of Project Chazon, little if anything proactive exists.

Nor do initiatives at a level to match the efforts of those with an antisemitic agenda – particularly in addressing fake images.

Likewise with university projects in the field, though they are developing increasingly sophisticated detection methods. Frequently vital research is not leveraged by Jewish legacy bodies or Jewish-run companies.

This is largely because these organisations and university departments prioritise their own focus, or do what their funding requires, often in a siloed fashion to protect organisational or commercial interests or independence in order to foster originality. It’s understandable. But it can mean a decreased cross-fertilisation of ideas for acting strategically against Jew-hate.

To address this, CAAI started a forum focused on how AI can tackle the mounting threat of online antisemitism. It brings together AI specialists, including heads of companies and academics, students, educators  and others. Jewish institutions, potential funders, those seeking collaborators, as well as coders looking for like-minded individuals for developing apps to combat Jew-hate, also have a dedicated place for information, support, or ideas from the field.

CAAI’s work is only one strand of a solution. Yet the Jewish community can now go further. Because just like Sir Keir stressed at UCL that there has never been a better moment to get involved with AI for British science and society, there has never been a better moment to build on the AI initiatives for dealing with Jew-hate already begun. Working tactically and, crucially, using the technology at scale, we have an unprecedented opportunity to push back against antisemitism.
 

.

Dr Jonathan​ Mye​rs is the ​founde​r of C​AA​I ​ jm​@​psych​o​nomic​s​.com ​