Transcript [00:00] People can express overt death wishes, or they can say below the line of an [00:06] article on George Soros, someone should give George Soros a shower. Right. And here you have the [00:12] word shower, combined with the name of a Jewish individual, which immediately activates the [00:17] picture of the gas chambers. Welcome to The Honest Take, the show [00:24] that goes past the headlines to find out what's actually true about Israel and about the people [00:30] covering it. I'm Ben Chertoff, and my guest today is Doctor Matthias Becker, the man who's training [00:36] A.I. to understand the anti-Semitic dog whistles we can't hear. [00:46] When most of us picture anti-Semitism online, we picture a swastika, a slur, or a guy in [00:53] a dark room. Becker will tell you that's the part that barely matters. He's not a political [01:00] scientist. He's a linguist. And for over a decade he's been reading the comments, not on [01:07] fringe platforms on the BBC. The Guardian, Le Monde His project decoding [01:13] anti-Semitism has gone through 300,000-plus comments across Britain, France and Germany. [01:20] And what he found 80% of anti-Semitism in those mainstream spaces never [01:27] uses a slur at all. It hides in a joke, a rhetorical question, a name dropped [01:34] like a password because the people writing it don't think of themselves as anti-Semites, which [01:40] means that nearly every statistic you've read about online hate is, in his words, built on [01:47] sand. He advises the State Department, the European Commission, the people building the AI meant to [01:54] catch this, and he was in the data on October 7th, watching something change he'd never seen [02:00] before. Doctor Matthias Becker. Welcome to The Honest Take. Thank you for having me. [02:08] So you're a linguist, not a sociologist. Not a political scientist. How does a linguist end [02:15] up running one of the most rigorous research projects we have on anti-Semitism online? [02:22] Well, this is a good question. I guess the correlation between linguistics and [02:28] discourse analysis and hate studies, um, is not really well known. Uh, most of the [02:35] projects are looking into digital hate, and conspiracy narratives are done by [02:42] political scientists, social scientists and so on. But of course, if you want to understand hate [02:48] speech and, you know, digital hate in particular, and conspiracy narratives and disinformation, of [02:55] course it is important to look into the pattern with which these these ideas are being [03:00] communicated. So my my journey actually really started with, um, racism [03:07] studies, uh, nationalism studies and how especially in, in Germany, postwar Germany and [03:14] until the present, how these ideas are being shared, especially in mainstream discourse. Um, [03:19] because in contrast to, uh, fringe communities, of course, neo-Nazis, white supremacy contexts, [03:26] these ideas, when they come up, are very often shared, um, through coded speech. Right, because it [03:33] would otherwise collide with the positive self images of a speaker. And from a linguistic [03:38] perspective, it's very interesting to see how these ideas are shared or communicated, what kind [03:44] of rhetorical devices you can find on the side of the speakers. Um, because, you know, you can of [03:50] course express an insult, a threat, a death wish. Incitement to violence. Overtly. Word by [03:57] word. Um. But you can also say. Say these things in in coded ways. You can imply them. [04:03] And and that's something that of course, applied linguistics are therefore to actually understand [04:10] these different communicative patterns, these rhetorical devices that are very often not picked [04:16] up by advocacy organizations, by, for instance, policy makers, [04:23] people who are concerned about the current developments, but you can see in public discourse [04:28] is a strong focus on on aggressive speech acts on, of course, on overt [04:35] glorification of violence. But this is just really a tiny part of the repertoire that you can find [04:41] online. You looked into 130,000 comments across the UK, [04:47] France and Germany in mainstream media content sections. Not These aren't dark web [04:54] forums that you're scraping for this. They're in the mainstream. Why would you expect [05:00] to find more there than in the sort of the sewers of the internet? Well, [05:07] the premise wouldn't be that we would find automatically more of these hate ideologies, right? [05:12] Or the or anti-Semitism in particular. And by now the dataset has grown to more than [05:19] 300,000 user comments. So it's an ongoing process. It's an ongoing journey into [05:26] into this world, which is, from my perspective, clearly under researched. Of course, there's a [05:32] whole range of projects dealing with digital hate, right? Especially when it comes to anti-Semitism, [05:37] especially after October 7th. The amount of of initiatives, um, grew [05:44] rapidly. Um, but I would say in the last 20 to 30 years if there [05:51] was an interest in current current forms of of hate speech. There was a strong emphasis on fringe [05:58] communities, especially white supremacy contexts. And I guess there's no surprise that you can find [06:04] all forms of all sorts of of anti-Semitism and different forms of racism on these [06:10] platforms. Um, but what is actually happening in mainstream discourse and [06:16] politically moderate online contexts? There is not much certainty about that, right? [06:23] Because it all starts and ends with the identification of a research design. [06:30] How do you try to understand something that is so complex and so fuzzy, [06:37] like digital hate in politically moderate contexts? And again, it is completely under [06:43] researched. And I think this is a huge problem because as soon as we, try to [06:49] understand anti-Semitism and anti-Semitism that becomes socially acceptable, permissible in in a [06:56] society like the American one, we need to understand what is acceptable, what seems to be [07:02] acceptable in mainstream discourse. Because if anti-Semitism, if the different forms of Jewish [07:07] hatred become predominant in those online spaces, we really have a problem. I guess fringe [07:13] communities will always have to deal with these kind of the, the, these obstacles that, for [07:20] instance, a a death wish coming from a neo-Nazi will be, you know, won't be able to [07:26] travel freely within the communication sphere without any sanctions, without any [07:33] clear forms of rejection. But as soon as it's becoming predominant in in mainstream discourse, [07:40] then we are living in times when Jew hatred becomes normalized. And this is a much bigger challenge, [07:47] and this is a much bigger threat to the American society. Definitely in different European contexts. [07:52] And that's why I thought this should be the focus. Not so much. White supremacy context. Not so much [07:58] neo-Nazis, not so much the radical left. But what can we find in the spectrum between conservative, [08:03] liberal and progressive online contexts? What is acceptable there? What kind of anti-Semitism and [08:09] what kind of rhetorical device can you see? Most people who don't have to deal [08:16] with this every day and who don't see this stuff every day. Most people would assume that [08:20] anti-Semitism online means somebody posting a swastika or, you know, calling somebody a slur like [08:27] you had mentioned before. What what does your data actually show, and what are you actually looking [08:32] for, the coded anti-Semitism, as you call it? Yes. Well, I would [08:39] say the the approach that we follow is always, um, uh, is not limited to to coded speech [08:46] itself. Right? We really want to understand the different antisemitic stereotypes and concepts in [08:52] general. So also Holocaust distortion, Israel-related anti-Semitism on the basis of certain [08:57] demonizing analogies. Um, speech acts like the glorification of violence that we could see after [09:04] October 7th. Um, and here, uh, the approach would always be [09:11] to identify these different concepts, these different ideas about Jews, and then to [09:17] see how these ideas are being shared. Is it a metaphor? Is it? Is it some kind of an indirect [09:23] speech act like a rhetorical question, for instance. Um, so people can, for instance, say Jews [09:29] exploit the Holocaust for financial gain, right? Or they can just ask a rhetorical question, for [09:35] instance, who's holding out their hand again? Uh, below the line of a of an article about the [09:40] culture of memory in Germany or Europe in general, people can express overt [09:47] death wishes, or they can say below the line of an article on George Soros, someone should give [09:53] George Soros a shower. Right. And here you have the word shower combined with the name of a Jewish [09:59] individual, which, if there is sufficient historical knowledge and awareness, [10:05] immediately activates the picture of the gas chambers. So you don't have a clear reference to [10:11] to death and to destruction. You can just use all sorts of allusions and puns [10:18] to to activate a hateful idea. In that case, this would be a coded death wish. And what we can see [10:24] when we, uh, when we went through all these comment sections in the context of mainstream news [10:31] outlets in Europe, and in this context, there was particularly the UK, Germany and France we were [10:38] focusing on, um, we could easily find, uh, like [10:44] average average values where, like 20 to 25%. Antisemitism below the line. Of these news [10:51] outlets or their articles. But then we could identify. Among these [10:58] antisemitic comments, we could identify a whole range of. Implicit examples. And [11:04] sometimes in the context of the BBC or The Guardian. If you could find like 20% [11:11] antisemitic tropes in the comment section. Sometimes more than 80% of these were [11:16] communicated in implicit ways. And that tells us something quite alarming, [11:23] which is that. All these statistical overviews that we have, all these claims that we now have. [11:29] Widespread anti-Semitism online is are very often built on sand, because they follow these [11:36] predefined understanding of how the idea of Jewish power or Holocaust distortion, or the [11:43] idea that Israel would be the new Nazi state, that all these things are referred to explicitly by [11:50] slurs or by statistically significant words. But of course, this is not the case when you do [11:57] qualitative close reading. When you go through the comment section and you see that there's a whole [12:03] range of words and phrases that can activate the idea of Israel being [12:10] a child murdering state. Right. Um, or uh, again, the idea that the Holocaust never [12:17] happened. You have the whole repertoire of human language to activate these ideas. So and there's [12:23] no, again, no statistically significant, um, realm of of of words that [12:30] over time come up. If you are interested in the apartheid analogy or in the claim that Israel [12:35] would, um, commit Genocide. Then I guess you have a high probability that you can [12:42] understand that the overall extent of these claims, because people usually refer to these [12:48] ideas by using these words. But of course, stereotypes are age old images [12:55] about Jews, usually negative character traits. And of course, you can claim these things about Jewish [13:02] deceit and hypocrisy and immorality in a million different ways. So what we need to [13:09] do is to to understand the variety, the complexity, the diversity of these, um, of [13:16] these statements and find ways to to measure [13:22] these references in the aggregate in order to say something substantial about, about [13:27] anti-Semitism. So I guess one of the main observations that we could, uh, make over the time [13:34] of the since 2020 and even before. I mean, I've been working in the field for more than 14 years. [13:41] Uh, is that the beginning or the the starting point of all these, [13:48] of this kind of research has to be an interdisciplinary, qualitative one so that you [13:54] understand, okay, this is for instance, um, a news article about October 7th or Bondi [14:01] Beach or, um, the news coverage about the decision a few weeks ago, um, that [14:08] you actually go through the digital aftermath qualitatively, qualitatively, with your knowledge [14:14] as a, as an expert, with your historical awareness, with your awareness about how language works and [14:21] to to classify what kind of tropes you can find and how are they communicated in [14:27] different ways through language, but also through imagery or multi multi multimodal ways. [14:34] So. And and as soon as you have a clear understanding of these discursive trends, you can [14:40] use that as the basis for, um, for training large language models or to, to, to [14:47] actually to get to the point that these models are actually able to replicate your decision so [14:51] that you can not only dig deep and understand the full extent of, of these [14:58] rhetorical devices, but that you can also say something representative through the scaling [15:04] process. And I think this is this is a really central observation. Antisemitism is a hate [15:10] ideology. And if you're interested in the presence of digital hate in politically moderate contexts, [15:17] you have to understand that people very often avoid, um, [15:23] explicit references. They because it would collide with their positive self images. So [15:30] therefore that is instrumental for your research design. You have to understand that this is the [15:36] thing. This is a pattern that you can find in different language communities in different [15:41] countries. As soon as you dive into politically moderate contexts. You have to see the diversity [15:47] of these statements. And then, um, you can actually say something substantial about the [15:54] real dimension of true hatred, uh, in those conversations. And a lot of this. You [16:01] mentioned training LLMs. Um, a lot of these coded words. Um, you know, [16:08] you also talked about the negative self-image and people wanting to avoid that. But I think some of [16:14] it must also just be to avoid our legacy forms of content moderation, which [16:20] relied on things like keywords. Um, so when you start to code it, [16:27] it gets through those content moderations. Is that right? Am I? Am I understanding that correctly? [16:34] Yes. Yes, absolutely. So I guess this definitely plays a major role. Um, [16:41] especially in the past when content moderation was, uh, more [16:47] dominant. Because now, as we all know, social media companies took a step back. [16:54] Um, and, uh, there are Community Notes, and, and it's kind of a based on a shared, uh, [17:01] knowledge about, like, what hate speech actually is that these interventions are actually happening. [17:06] And of course, you have the, the, the system of trusted flaggers, um, uh, who, who give [17:13] recommendations about like what should be taken down. But in general, this kind of regulation is [17:19] definitely not as prevalent as it should be. But I would also argue that, uh, content [17:25] moderation is just one factor among many, right? It's something that we can definitely identify, [17:32] even in fringe communities in in the context of white supremacy platforms, it's also a high [17:39] presence of these coded statements. And why is that? There's basically [17:45] no content moderation whatsoever in those online spaces. So the reason why why you can find [17:52] these, um, coded forms of, of, of digital hate and conspiratorial thinking [17:59] on those platforms is because it's attractive, because it is attractive to say something [18:06] like, you know, we allude to the idea that Jews would control the White House or Hollywood, but [18:11] not, again, not a word not explicitly, but through certain [18:18] innuendo or through metaphors that not everyone understands. Right. It's a little bit of the whole [18:24] thing. When you think of the Matrix movie, The blue and the Red pill. If you understand [18:30] what is really. If you are able to understand what is really happening in the world and you want to [18:35] share it with your own peer group, and you want to do it in a in a compelling way, you use some kind [18:41] of a, um, a communicative function that not everyone around you understands [18:48] because that builds a stronger like, um, communicative proximity between yourself and [18:55] the peer group. You want to reach out to. And at the same time, you can imply that the society [19:02] around you doesn't get it right. So this is also a way of, um — of course, dog whistles [19:09] would be another would be another term for that, that you just create a stronger bond between [19:15] yourself and the people you want to, you want to reach with your message. So this is a very [19:22] central aspect. Content moderation is definitely part of these trends. And of course, in offline [19:28] context, there was always the, the a high probability that you you might get sanctioned if [19:34] you spread hateful ideas. But, um, it is just one side of the story. [19:41] And this is something that we need to understand. And if we want to understand how these things [19:45] evolve over time, we need to build systems that can, um, [19:51] understand these allusions, these jokes, these different metaphors, metaphorical phrases, um, the [19:58] way they appear in, in conversations. I want to get into the data because you [20:05] have been studying this for years now. Much of your research was before October 7th. Can you [20:12] walk me through what you saw? The difference between, uh, what how this online anti-Semitism [20:19] presented itself before October 7th and then what you saw real time, uh, During and after. What [20:25] changed? Yes. Um, yeah. This was definitely a turning point. [20:32] October 7th. Um, so I've been I've been interested in Israeli related antisemitism for quite some [20:38] time now. I also did my PhD about this, um, a comparison between Germany and the UK and, um, [20:47] uh, what we could see already in 2012. Uh, autumn 2012. Summer [20:53] 2014, May 2021. All of these were escalation phases in the Middle East conflict. Right. And [20:59] something that you could observe over the years was as soon as there was an escalation in the [21:05] conflict within the conflict, um, stereotypes and, and, uh, demonizing analogies like the [21:11] Nazi analogy, uh, went up, went through the roof. Uh, there was a clearer, uh, correlation between the [21:18] two sides. And so it was not surprising to see, you know, for instance, in summer 2014, that. For [21:24] instance, allegations of Israelis trying to find a the the the [21:32] final solution, um, of the Palestinian question and phrases like that, we're just very, very [21:38] present in, in online communication. Um, on October 7th, [21:46] that has changed completely. So it was not so much the projection of stereotypes [21:53] and analogies onto the Jewish state, uh, onto the Jewish outgroup. But these were there [22:00] were a lot of forms of so-called self positioning. So people weren't actually saying, you know, Jews [22:06] are evil, and that's why I hate them, or that's why I'm in favor of the destruction of the Jewish [22:13] state. People started or reacted to the pictures on that day, to the to the, uh, clips and [22:20] images that were circulating by, um, glorifying, glorifying like what they saw. [22:27] So there was an open celebration of, uh, of that kind of violence and murder and slaughter [22:34] and in a way that I've never seen before. So from moving away from projections [22:40] towards, um, self positioning, towards an open affirmation [22:47] of, uh, the cruelty on that day, and that's something that, that could be that was echoed by [22:54] various research projects, not only mine, we we published a, a case study two weeks after October [23:01] 7th. I think it was the first study that came out about the online reactions. And the study was [23:06] called, uh, it's called Celebrating Terror. And that kind of is, in a nutshell, what we could see that, [23:12] um, again, it was not the projective element that is so central in the context of [23:18] anti-Semitism as a hate ideology. But it was it was statements like. I'm I'm so happy about what [23:25] I see. This should happen from now on in the next 60 years. Right. And maybe afterwards you could [23:32] actually see these stereotypes and analogies, analogies coming along. But it was clearly the [23:39] brutality of, uh, the reports coming from southern Israel that triggered, [23:46] uh, an open and open celebration among, uh, web users at that definitely [23:52] stayed part of the even of mainstream discourse over the course of the following weeks. Over time, [23:59] as soon as the military operation of the IDF started in Gaza, uh, [24:06] you could see that the old projections returned. But the aspect of, of like affirmation, [24:13] celebration, glorification of violence, the justification, of course, that really remained part [24:20] of even politically moderate contexts, and something else that we could [24:24] observe was the overlap between anti-Semitism and misogyny and affirmation. Of. Gender [24:31] based violence. Right. If you think of the female festival goers, and it doesn't matter if we look [24:37] into the comment sections in France, in the UK or in Germany, which again at that time was the the [24:44] focus of my work of the Decoding Antisemitism project. Um, there was a clear overlap between [24:50] anti-Semitism and misogyny in a way that I've never seen before. The statements were not just [24:55] saying, um, these female festival goers deserved what they you know, what, [25:02] they got a day. Um, but it was even to some extent, um, pornographic, [25:09] um, uh, verbiage that you could find in the comment sections. And, um, this is something that [25:16] was quite new, something unexpected for my research team at that time, [25:23] especially since antisemitism and mainstream discourse very often doesn't overlap with other [25:30] hate ideologies. Quite often, you know, there's an anti-Semitic statement coming from someone, and [25:36] there's counter speech from someone who has a problem with anti-Semitism, but might at the same [25:41] time reproduce racist stereotypes. So you have this kind of juxtaposition right between these [25:46] two sides. It's more kind of a phenomenon of right wing fringe discourse that [25:53] anti-Semitism and racism overlap. For instance, when you think of a great replacement theory [25:58] where these two belief systems come together and are intertwined, um, [26:05] but usually anti-Semitism as a hate ideology stands on its own feet in mainstream discourse. [26:11] And that definitely changed on that day. After October 7th, you started doing what you [26:18] call event studies, these real time analyzes after a major violent event. [26:25] Can you walk me through what you found? After studying Charlie, the Charlie Kirk killing, [26:32] the Capital Jewish Museum shooting, and then the White House Correspondents Dinner attack? [26:40] Yes, certainly. Uh, what we could see, especially on, uh, everything related to Charlie Kirk, [26:47] was, uh, this was a quite interesting case study because, um, there was a [26:53] strong also there, strong affirmation of political violence on that [27:00] day. Um, so everything that we could identify in [27:07] previous case studies, because case studies were always some, some like a central part of of our [27:12] work. Because if you acknowledge that you can only say something substantial about anti-Semitism, all [27:19] its forms, that you need to have a fine grained, granular approach to it, [27:26] the best way you can you can actually approach the topic of digital hate is by identifying [27:32] something like a discourse event, and then look at the online reactions that followed. So this is [27:39] definitely a very helpful approach to make good use of your time as a [27:45] researcher and to understand how different, um, different discourse events [27:52] trigger different forms of anti-Semitism. So on that day, um, the assassination of Charlie Kirk, [27:59] uh, there was clearly, uh, an abundance of, um, affirmation of political violence [28:06] in combination, of course, with conspiracy narratives. uh, I guess the most famous example for [28:13] conspiratorial thinking in that regard to this discourse event is, um, Candace Owens, [28:19] but not only her and her followers and, and, um, influencers that are [28:26] connected with her. Um, even outside that realm. You could just [28:33] see a really a high like an omnipresence of of of of these conspiratorial [28:39] ideas after that. As you mentioned, we conducted a range of studies. Of course, Bondi [28:46] Beach was always a big topic. Um, we, uh, had the White House, uh, dinner [28:53] shooting. Um, Dan Bilzerian and I mentioned him before all these case studies. Um, have something [29:00] in common. They emphasize the the the high presence of of these conspiracy theories [29:06] in mainstream Discourse. There's also some really interesting observation. There was, uh, last summer, [29:13] the the the conflict between Iran and Israel escalated. Uh, Israel was the dominant, um, [29:20] uh, factor when we think of the US Israeli relations at the time, um, getting, getting [29:26] engaged in, in, in, in this kind of war, uh, against [29:33] Iran. And of course, when you look at the online reactions during these days, there was a whole [29:39] bunch of demonization of Israel, um, in conservative but also in progressive [29:46] contexts and liberal contexts. There was a clear focus on American news outlets and [29:53] their followers. And when you compare these results with the online [30:00] reactions to, um, February-March this year, when actually the US [30:06] starting to be involved in that war. You could see that there was a clear shift from [30:13] these standard, like demonizing patterns directed against the Jewish state towards [30:19] conspiratorial thinking, which is, you know, understandable, which makes perfect sense. Uh, [30:26] given that from from that moment onwards, uh, America played, like, a very, very [30:33] active role, a very central role in that war. And so you can you can actually [30:40] understand that kind of shift from standard demonization towards something that is still [30:46] compatible with, with the intent to, to, to [30:53] demonize Jews, but at the same time to make sure that, um, [30:59] it's not per se directed against your own country as an American. Right? So if America if the US [31:07] is doing something that again collides with a positive self-image. Um, the only explanation [31:13] you can, you can push is that this is happening because of the orchestration of an external [31:20] state in this regard. This would be Israel. So and that's, that's the really the interesting part of [31:26] it. I already mentioned the comparison between, um, the different escalation phases in the Middle [31:32] East conflict and how, you know, in different years, the repertoire of demonization shifts. [31:39] Um, this whole conflict with Iran since last summer tells [31:46] you the exact same story. The geopolitical constellations change, um, the way in which, [31:53] uh, mainstream media report on, on these, on these events might have shifted as well. And then you [32:00] can see a clear effect on the, uh, on the on the [32:07] reactions. You can see in, in on that debate culture and that actually makes this kind [32:14] of fine grained, um, granular linguistic analysis so, [32:21] so powerful because you see, you see the context, you see a discourse event, you see the you [32:28] you try to comprehend the reactions that you can find among web users, and that helps you [32:34] understand what is actually driving your own society. Right? Because language on digital [32:40] communication always echoes, uh, collective attitudes. So if there's a if there's a [32:47] certain shift in those, in those attitudes, you can reconstruct these shifts by looking at [32:53] the, um, predominant patterns of communication in online discourse. Online discourse is obviously the [32:59] marketplace of ideas these days. This is where things are happening. I would be very cautious [33:05] when it comes to surveys that try to understand the presence of different ideologies, because [33:10] again, we have factors of social desirability. We have priming effects. Surveys [33:18] represent highly artificial situations where people just, you know, get get it. You're talking [33:23] about surveys where somebody calls up and says, what do you think about the Jews? Somebody is not [33:28] likely to answer that as honestly as a wink and a nod on a comment section anonymously? Yes. Yes, [33:35] exactly. Even if it's, you know, if even if it's, um, if it's I mean, in general, [33:42] these are anonymous, um, and anonymous contexts, right? There's there's no question about that. [33:49] But, you know, looking like knowing that it is some kind of a cultural [33:55] norm to, to reject, Um. Uh. Prejudice. Hateful [34:02] thinking. Uh, generalizations. Essentializations. All these things are clearly, [34:09] uh, within a certain percentage among, uh, within a society are clearly rejected [34:16] phenomena. So if you ask these questions and it's not about, like, perceptions of a certain hate [34:23] ideology within a minority, there you can be certain that the responses are pretty [34:28] straightforward and and are authentic. But if you ask the majority of society what they really [34:33] think about this or that minority, the probability probability is rather high that, uh, the results [34:40] are pretty distorted. And because it is, again, this kind of artificial situation. [34:47] And even though you are anonymous, they are still factors that determine your reaction. Like [34:54] again social desirability. So therefore I would always take these results with a grain of salt. I [35:00] would always make sure, um, to emphasize if if you want to, if [35:07] you want to understand what is driving your own society, what is happening at the moment, [35:13] you you get far more reliable results if you analyze self-motivated statements. [35:20] And in that case, digital communication, online discourse is self-motivated. So you have this kind [35:26] of discourse event. People are looking at some kind of media coverage, and then they want to [35:31] position themselves to that problem, and they use it. And they they do this in very particular ways. [35:38] Right? They, they they sometimes do it very overtly and sometimes they use coded which uh, or [35:44] multimodal patterns. Um, and the reasons for this or that choice can be reconstructed over time, but [35:51] it's not the primary thing. The primary interest on our side would be if there is this discourse [35:58] event in this specific measurement period, what are the the concepts of [36:04] stereotypes, um, with which web users react to that? You mentioned the [36:10] misogyny in and the overlap of misogyny and anti-Semitism, and it just makes me think [36:17] of the the term manosphere and the anti-Semitism that seems baked into so many of those [36:23] influencers. You had, uh, analyzed the Dan Bilzerian announcement that he's running for [36:30] Congress in Florida, and he ran a study on the comment sections there under that coverage. What [36:36] did what did you find? So something that was really, really [36:42] interesting in this context was, um, that Dan Bilzerian, you know, the so-called [36:49] king of Instagram or former king of Instagram, um, That his ideas were [36:57] very, very welcome among a different target audiences. Right. He was invited. He was invited by [37:04] George Galloway, a British politician, clearly a left-wing, uh, [37:10] anti-imperialist, uh, politician or public figure. Uh, and of course, at the same [37:17] time he was invited by, by other hosts who you might identify with with the manosphere. Um, [37:24] or clearly like right wing influencers and, um, it was there [37:31] was a clearer, um, uh, appreciation for what Bilzerian [37:38] had to say about, like, what is actually happening right now here in the US. And one central aspect [37:44] was, of course, that he thinks that there is something like Jewish supremacy that is the [37:49] biggest threat for American society, for the world in general. Um, he, uh, [37:56] challenged or he's challenging, um, another, uh, politician, a Jewish politician [38:02] that he very often called, um, uh, “fat Jew”. Um, different forms of [38:09] demonization and conspiratorial thinking, uh, came from him. And [38:16] throughout all these comments sections that we analyzed, uh, in each case around 3,000 comments, we could see [38:23] that there was strong appreciation for this kind of, uh, scenario, right, for this kind of, [38:30] um, defamation and, uh, of course, engagement in, in digital [38:36] anti-Semitism, the overall amount of, of and of confirmation or [38:42] even just reproduction of anti-Semitic tropes, in particular conspiracy theories was, uh, [38:49] really mind boggling. So even the smallest, uh, the lowest level of [38:56] anti-Semitism in that context was still much higher than the average amount of anti-Semitism [39:02] we could find in the context of other discourse events. So here you could see that you don't even [39:08] need a specific event. For instance, an anti-Semitic incident, uh, somewhere in Brooklyn or [39:15] in the streets of London or whatever you're interested in, it is just some kind of a statement [39:22] that has been pushed by a public figure like Dan Bilzerian [39:28] that leads to a huge amount of of, um, uh, anti-Semitic rhetoric in the [39:35] comment sections. And it's interesting that, like, one of the most interesting observations [39:42] was that in one of these channels that we analyzed, uh, one of these YouTube channels, There [39:48] was a clear rejection of their positions. Right? There was there was a clear [39:55] attempt to say, you can't call your your opponent the “fat Jew” because this represents anti-Semitism. [40:02] And below the line of that clip, there was the highest amount of anti-Semitic [40:09] rhetoric, more than 50%. Wow. And that tells us something that, um, [40:16] is very often not really well understood, um, [40:23] which is the responsibility, the responsibility, uh, like from coming from a, from on the [40:30] side of the host, what that person should actually bring to attention. Right. I remember there was a [40:35] big debate a few a few years ago between, uh, Lex Fridman and Sam Harris [40:42] about, um, the fact that Fridman invited Kanye West to his show. And Sam [40:49] Harris was I remember well. Very, very straightforward with his critique and said, maybe [40:55] it's not a good idea to give someone with these ideas or his affirmation appreciation for Adolf [41:01] Hitler to give that person a microphone. Right. It doesn't really matter how you, as a host, uh, [41:08] try to dismantle all these ideas because you, you still you you you start like, [41:15] being engaged in something that you cannot really control because you give that person the [41:19] opportunity to share their viewpoints, and therefore, you become an enabler of these [41:26] normalized trends in even a mainstream discourse. And I would say, I'm not sure if Lex Fridman [41:32] actually understood that critique at that time, but in this case study that we conducted was a [41:39] proof for the fact that even if you if you interrupt your podcast [41:45] guest. After each sentence, we say and emphasize that this is actually an expression of an age old [41:51] hate ideology. It's not always clear. It's not. You cannot guarantee that the message, um, [41:59] is getting to the target that the person had in mind. Right. Because it is it is a language. And [42:06] communication in general is very often working unconsciously. Or, you know, when we think of [42:11] imagery, for instance, if you if you stand in front of a school class and you, you want to, you know, [42:16] you want to criticize or you want to you want to deconstruct what anti-Semitic propaganda is, and [42:22] you show a couple of cartoons that you say, this is bad because X, Y, and Z. Um, you cannot [42:29] guarantee that this is really the message your students go out with, uh, because images work [42:36] on their own way. Their own way. They work unconsciously. The same was true with with with [42:41] language in general. um, if you just give that person a microphone and the opportunity to share [42:48] these ideas, then, um, you, you, you participate in this whole [42:55] arrangement. Of course, I wouldn't say that. I'm against the idea of, like, [43:01] presenting hateful ideas and proclaiming them and deconstruct them and so on, but it has to [43:08] happen in a very, I would say, in a very moderate and a very controlled, uh, scenario. Otherwise, you [43:15] just have it just leads to a lot of really, um, uh, [43:21] negative effects on, on, on our society. And our society today is very much like, framed [43:28] or determined by these, by this kind of completely unlimited discourse that we can [43:35] see right through social media, through the communication conditions that you have online, [43:40] that people can actually say whatever they want. There's no accountability. And of course, slowly [43:46] but steadily, certain ideas become more and more predominant because there are actually no [43:51] gatekeepers. So anyone can say whatever they want. And and language and discourse, [43:58] of course, lead to, um, to attitudes, to sentiment. That's you know, the history [44:05] of Germany shows shows this to us. Antisemitism was a normal part of everyday culture. So if it's [44:12] already part of a discourse, it's only a tiny step to physical violence, to institutionalized [44:17] violence. And so this whole topic of like, how do we [44:24] actually draw the line? What should be part of the conversation? Um, I'm not sure if [44:31] anything that is out there, uh, should just be presented in, in one way or [44:37] another with the headline. Sunlight is the best disinfectant, right? Because if we talk about this, [44:43] we can dismantle it. We get rid of that idea. I think a language and the idea of ideology at the [44:49] ideological basis for specific forms of discourse, uh, cannot [44:56] be, um, controlled in any possible. Like every minute. Right. I think we [45:03] should we have a responsibility for what we actually present to our audience. And, um, [45:09] and, uh, if, if we, if we invite people like Alex Jones or Kanye [45:16] West or in that case, back then, um, it isn't it is again, it is not always [45:23] clear that we actually achieve our goals by just confronting the person with the truth or with, um, [45:29] a well-balanced opinion, because the narratives, the more aggressively the narrative is, the more, [45:36] uh, probable is the success of the person that utters these ideas. And also the. [45:43] It's such a it's such a common move of these conspiratorial people [45:50] like Alex Jones is if I'm getting pushback, that means I'm right. And it seems to be, you know, one [45:57] of the things that they take to prop up their argument is that the more flack they get, the the [46:02] more they're over the target. I think, is is the Alex Jones term about that? The pushback makes it [46:09] worse. Is what you found. The the the amount of of of anti-Semitism was definitely higher [46:16] in, in that particular episode than in those cases where people were just, um, you know, appreciating [46:23] the, the, the courage of Bilzerian. So that's a very interesting empirical observation. [46:31] The, um, you know, a lot of conversations with my wife, for instance. [46:37] We will comfort ourselves by saying, well, a lot of this is just online discourse and people online [46:44] are crazy. How much correlation is there, do you think, between the online discourse and [46:50] offline action? Wow. Yeah, this is a very good question. And so far [46:57] nobody can actually like reliably answer that question. Of course, there are attempts to, you know, [47:03] then if you think of, um, the, uh, the shootings, for instance, in Pittsburgh or [47:10] in Germany, in Halle, or Christchurch, that after these, um, after these crimes, [47:17] people were looking into the to the statements of the perpetrators. Right. And, and the manifestos and [47:24] so on. The same was true with Washington, DC in 2025. Um, so there are [47:31] attempts to, to to look at the correlations between the two. But at the moment we still don't [47:36] have absolute certainty about these. Like how, how how much kind of overlap you [47:43] can identify. Because for that you would always need to have, um, a very [47:49] precise bird's-eye view on the reality in online discourse, in online communication. And if a model, if [47:55] you still don't have that at the moment when we talk about social media studies, digital hate [48:00] studies, you name it, it is um, most of the publications are [48:08] clearly exploratory because either you just focus on a specific online environment or, you know, [48:15] specific influencers. And through reactions, these influencers got, um, or you try to cover [48:22] this as much data as possible, but then your understanding of verbal violence and [48:29] conspiratorial thinking, uh, is rather superficial because the bigger your dataset [48:35] gets. The more limited are your your your tools to actually comprehend the different patterns. Right. [48:41] So so usually most of the most of the studies or case studies in the field of [48:47] digital hate research are focused on slurs. Right. As I mentioned at the beginning. And so slogans [48:54] that always that always repeat themselves. But for, for people [49:00] who who are actually planning something like an attack on a, for instance, on a synagogue, um, [49:08] um, there of course, the statements coming from, from these individuals need to need to [49:15] be analyzed in more detail. Right. And you need to take into account especially the aspect of, again, [49:21] conspiracy theories, because conspiracy theories in themselves are far more [49:28] alarming than just Expressing demonizing stereotypes about Jews, for instance, because [49:35] stereotypes are very often based on the idea, okay, these people are, uh, inferior or [49:42] they are just, you know, they have no moral compass, you know, clearly constructions of [49:48] difference. Whereas conspiracy conspiracy theories always imply something like, [49:56] um, we would call a self victimization, because when you think of a perpetrator in Pittsburgh, [50:03] you know, he was completely hijacked by the great replacement theory and the Great replacement [50:10] theory, just as other antisemitic conspiracy theories are based on the idea that there [50:17] is some kind of a Jewish lobby, uh. Powerful lobby [50:24] that operates in the dark behind the curtains. And you don't know what will happen next to you, to [50:30] your family, to your to your respective in-group. So the aspect of self victimization is extremely [50:36] central for, um, conspiracy narratives and of course connected to these [50:43] ideas of, of like. That you and you are kind of represent [50:49] victims. Um, are emotions like fear, anxiety. Right. You are not. It's not so much based on [50:56] hate, on on hate or disdain and related emotions, but actually fear that you might [51:03] actually lose everything. And as soon as a collective starts, um, like being hijacked [51:10] by the emotions, the probability of physical violence is going up. Right. [51:17] So, so looking at conspiracy theories and all and of course, people who think that the great [51:24] replacement theory is true or, um, I don't know, for instance, that, um, [51:31] uh, uh, Charlie Kirk was killed by the Mossad, for instance. And, you know, you don't know what's going [51:38] to happen next. Quite interesting. Also in our case study about Dan Bilzerian. Uh, they were statements [51:43] like, uh, remember that Dan is not suicidal. Right. Which is a very [51:50] interesting way of, like, alluding to the fact that, um, he might be dead tomorrow and we all know [51:57] who is responsible for that. He's not suicidal. And and so if this is the scenario you as [52:04] a web user draw every day, or you get confronted with every day, you think yourself like living in [52:10] a, in a, in a society that is controlled by something that you cannot see, that you cannot [52:15] understand. And of course, the the next step is fear and, um, lack of control. [52:22] And of course, we need to get our control back. Right. And so this shows pretty well how [52:27] conspiratorial thinking is. As soon as it's actually getting normalized in various online [52:33] environments and can actually lead to real world harm and violence in a [52:40] way that standard stereotypes, prejudices, uh, cannot. And and so I [52:46] think I mean, I only partially answered your question. I think it's, it's it's very important to [52:52] understand that the, um, that our instruments, our tools to, to understand [52:59] digital hate and all its forms are much better than two years ago, thanks to the technological [53:06] possibilities we have today. But of course, the bottleneck is still can these models, these [53:12] classification models understand these different patterns of hate? Are they actually able to [53:18] identify the stereotype of, um, Jewish power? And how do [53:25] they draw the line between criticism of lobbying versus, [53:31] again, conspiracy theories about Jewish organizations? All right. And then models need to [53:37] learn that. And for that you need expert annotation. That brings us back to my project and [53:42] projects from from other research centers. It is absolutely important that you test these models, [53:47] that they perform well. And as soon as they are able to do that, we could actually answer these [53:52] questions about the correlation between online hate speech and offline violence. Before that, [53:59] it is very often just an exploratory attempt to to understand these correlations. These correlations [54:06] are certainly there, but we are definitely not yet at that point [54:13] to, to predict, uh, this kind of violence that very often comes [54:19] from lone wolves. Right. I wrote an article about the issue that a lot of law enforcement agencies [54:26] are pretty much focused on aspects of, or approaches like social network analysis, trying to [54:32] understand how different actors, radicalized actors, extremists, potentially terrorists, even [54:40] how they, um, how they prepare themselves for, uh, [54:47] certain crimes. And this is very often based on the way in which these different actors cooperate, [54:53] right? How they actually how they, you know, interact online. Right. And how much they are [54:58] actually organized in certain groups, etc.. But in times when a lot of people who, [55:06] uh, engage in this kind of violence get radicalized in the first place within their [55:12] online community, within their peer group, and they just read every day what I said before they read [55:17] every day about, um, their country going down because of some kind of lobby. Um, and they [55:24] just take all that in until it's kind of accumulating to a, like a comprehensive worldview [55:31] where only one thing can help us get out of this vicious cycle. Uh, the only [55:38] thing you can actually anticipate that something bad is happening, that this person is planning [55:43] something atrocious, um, is based on, on the way this [55:50] person expresses themselves online. Right. It's a language that actually tells you, okay, this person [55:56] is completely like drowning in conspiratorial thinking. Therefore, the probability is high that [56:02] something could happen. You never know 100%. But it's the language, the discourse itself, that could [56:08] shed some light and help us predict, uh, future forms of offline violence that we. [56:15] We cannot even comprehend this in this moment. You. I want to get into some of the [56:22] solution here. AI content moderation is supposedly advancing very quickly. You just alluded to [56:28] that. You. If I'm not mistaken, you are advising some of the larger [56:36] tech companies. Also the US State Department. Um, what [56:43] do you tell them when they ask you what we should do differently? What what are you telling these [56:51] companies and governments? So, um, first of all, these advisory [56:57] activities are clearly, uh, they they definitely happened at a certain time frame. At the moment, uh, [57:04] I'm mainly focused on, uh, advising, uh, different research centers. Uh, so we don't have ongoing [57:11] collaboration with social media companies. What we are actually trying to establish these days is a [57:16] closer link to AI companies. Um, in order to help them improve their, uh, [57:23] their guardrails of their models. Um, because very often, uh, domain expertise is not really part of. [57:30] Are they open to it? Because I know that, you know, Wikipedia is a stumbling block and Reddit is a [57:36] stumbling block because they feed these large language models their, their information. Um, so [57:42] yeah, I'm interested are the, you know, is Anthropic not to name names? Are are the AI [57:49] companies in general? Have you. Have you gotten any positive feeling from them about this? I can't say [57:55] much about this at this point, but I will definitely, uh, hopefully at our next conversation. [58:01] But it's definitely, I would say just from a very naive perspective, it is, um, I think the chances [58:08] are much better to, to actually move in a good direction than, uh, with the conversations [58:15] we've had with social media companies. You know, I can't generalize at this point, but I would say, [58:21] given that in social media discourse, a lot of engagement is actually really coming from [58:28] like pushing certain like negative emotions. This is, I would say, not really [58:34] an incentive or the business model from, from AI companies on the side of AI companies. So I [58:41] think there's definitely a higher probability to move in a good direction when when it comes to [58:47] other research centers, um, the thing or the approach that we follow is very [58:54] often to, to to think about digital hate in new ways. For instance, I'm working very closely with [59:01] AddressHate, a think tank here in New York, um, that uh, tries to, [59:07] um, strengthen the, the, the connection between the different research centers in the US, but also in [59:13] North America, in Europe to bring actually these different researchers together who look, for [59:18] instance, into discourse like what I do, or people who are interested in malicious actors [59:25] right, about bot farms and orchestration, and those organizations or research centers who are [59:31] interested in algorithmic amplification, to actually have an ongoing dialog between these [59:37] different parties, between these different centers of expertise. Because this is a guess, uh, I [59:44] would think the the basis for a research community that is able to understand [59:52] these current trends on, on various levels and inform the different target groups like [59:58] education, like policymakers, like legal experts platforms, but also law enforcement [60:05] agencies. We have to understand that, again, human beings communicate these Exclusionary [60:11] ideas in a million different ways, and we need to accept the fact that it takes time [60:18] to generate training data, to build data sets that are helpful for creating [60:26] external knowledge bases for large language models, so that these models over time understand, [60:32] um, the different again, the different layers, the different communicative patterns of digital hate [60:39] so that they can replicate our decisions. That's that's I guess, the missing part. Um, so today [60:47] we have opportunities to get raw data in different ways, right through social listening [60:52] companies and so on. On the other hand, we have we are technologically extremely advanced. We have [60:59] large language models that understand irony to some extent, that understand jokes, that understand [61:05] all sorts of rhetorical devices that could have never been covered a few years ago with a high [61:11] precision. But what is definitely missing is this middle part that you bring in experts [61:17] from, like experts on these different belief systems from a historical and [61:24] contemporary standpoint. People who work in communication science, who work in psychology, who [61:29] work in media studies, who actually take all their theoretical [61:36] knowledge and, um, apply it to to [61:43] online discourse and over time to, to inform or to, to, to, to give, to [61:50] build some kind of a basis, the foundation for the models to understand how these different belief [61:56] systems are being shared. This is the most important part and this is at the moment the [62:02] weakest part in the whole pipeline. You spend your days in comment sections reading hatred [62:09] and quantifying it and coding it. And you've been doing this for years. How do you keep [62:15] yourself from going down a terrible rabbit hole of despair? Oh, this is a [62:22] this is a very good question. I think at the beginning, um, yeah, more than ten years ago, it was [62:29] pretty hard. It was pretty difficult because at that time I was very much focused on on extremist [62:36] content. And of course, this has like really a clear effect on you as a researcher. Right? This is [62:42] a this is of course, there are coping mechanisms and people get used to it or, you know, which is, uh, [62:50] they just manage and maybe create some kind of a cynicism. And fortunately, [62:57] this is not a destiny that I share. Um, I think it was pretty helpful for me to [63:04] move away a little bit from the very like extreme forms of radicalized online environments, [63:11] and to focus a little bit more on mainstream discourse, because if you even though the [63:16] underlying concepts, the stereotypes are basically the same, just the rhetorical, the communicative [63:23] appearance, like through the implicitness that I mentioned before, um, helps you, [63:30] uh, deal with these, these phenomena. But at the same time, it is also just the [63:36] observation that you actually learn a lot about your own society or the society. You, you, [63:44] you do your research on. Right. It's it's again, it's a social media studies if you do it right. If [63:49] you bring in the right disciplinary frameworks, um, social media studies can [63:56] really be some kind of a magnifying glass for, um, attitudes and [64:03] sentiment and overall dynamics within. Within a language community. Within. Within a. [64:10] Within a country. Um, this is something that I, I think is extremely [64:17] insightful. Also like just looking at anti-Semitism, right. Anti-Semitism is not about [64:22] Jews. It's always about the anti-Semites. It's always about what what is going on in in a [64:29] society like the American one these days, when there seems to be kind of a complete like lack of [64:35] orientation on the political level, or the question about the war on Iran and all these [64:42] projections that you can find online. I think this is always the basis for understanding the, um, [64:49] the current state of American democracy. And, you know, obviously there are lots of [64:56] correlations between anti-Semitism and anti-democratic resentment. Right? It's in Germany, [65:02] in a country like Germany. It is. It goes without saying. If you if you if you study antisemitism, [65:07] you also move. You take like a huge leap forward in like understanding [65:13] the current state of democracy. Right? Because antisemitism is clearly connected to ideas of [65:19] anti-elitism, that you don't believe that this institution is doing something that is good for [65:24] the community, for for the society as a whole. So you reject certain institutions, and over time you [65:30] actually have a problem with democracy in itself. And so these correlations are there. I think [65:36] therefore, um, studying, uh, digital hate and antisemitism in particular [65:44] helps you understand where, where your own society stands at the moment. And that's something that is [65:50] very interesting. It's again, it's quite psychological. It's quite an interesting [65:56] observation of, of the Zeitgeist. Um, and it has [66:03] major implications because in order to tackle it, you need to understand how it operates. It's like, [66:08] you know, it's like a virologist in a laboratory. You have to understand how the virus operates, and [66:13] then you can build your, uh, based on that, your your remedy. Um, that [66:20] exactly is the case for ideologies. If you just let it grow, uh, it will flourish and it will [66:27] normalize. But if you understand how why, for instance, young people who in other contexts are [66:33] very open minded, are moderate, are progressive, etc., and you are completely against racism, why [66:39] they get seduced and hijacked by certain ideas when you see that in real time online, how it [66:45] happens through what kind of, um, argumentative patterns, through what kind of [66:51] analogies you can use that in your in your intervention, in your inoculation. Because then [66:58] you see why they are, um, why they are open to that. And this, of course, [67:06] I think there, there are just no limitations in like dealing with these issues. People are [67:12] terrified of AI, the possibilities. You know, they're the AI doomers out there. Um. [67:20] And there's been a lot of talk about how, you know, Grok turned itself into, uh, Hitler loving, [67:27] uh, whatever. Do you think with the right training, there's a way to AI [67:34] could actually come in and help really help us understand this to a point where maybe we can [67:41] make some progress. Finally. Absolutely, absolutely. It's it's completely [67:47] I understand, uh, the the the the fear and the uncertainty about AI and I [67:54] completely, Uh, I can completely see these concerns. Absolutely. [68:01] But at the same time, you know, this is not the first time when human beings talk about [68:06] technology and the effect of technology in a, in a positive or in a negative way. It really [68:13] depends on the intent, right? It really depends on what you want to do with that. And of course, AI [68:19] can be used and misused. Uh, for the, uh, normalization of anti-Semitism, just [68:26] like any other hate ideology. And, um, and we could easily get to a point that, uh, you [68:33] know, social media cannot be used anymore because there's just such an abundance of distorted, [68:40] hateful, um, you know, materials and obviously deepfakes and so on. This is [68:47] all, you know, very like, consistently discussed also in public discourse. And of course, [68:53] now, you know, a few years ago you needed to have a whole team in order to push a certain narrative [68:59] and maybe to orchestrate, um, a public discourse in some way. It was extremely [69:06] expensive, and it was extremely tedious to to get to something. And now you just need one person [69:12] with the right skill set, with the right models. And, um, and you can come up with a [69:18] whole, with a whole system of, of, um, like a whole waves of [69:25] propaganda about, like, what Jews would represent. Right. And there are a range of studies that came [69:29] out just in the last months about the 49 “rabbis”, etc.. I mean, this was pretty much in the [69:36] media, and then people understood that this is all kind of fake, right? But it looked so authentic. And [69:42] people who just scroll through the web and who have kind of a divided attention, they just [69:47] consume it to some extent unconsciously, and it becomes part of their worldview. So this is all [69:53] real, right? This is a huge problem. And again, talking about deepfakes that are sometimes so [69:58] sophisticated, um, that even experts struggle, um, identifying the, the, the [70:05] these, um, the distorted dimension of it and and that this cannot be reality. Um, this is [70:12] all well known, but at the same time, there's just no way around around it, right? There's no way that [70:19] we can understand, but we can make sense of current society discourse without the help of AI [70:25] because, um, uh, meaning is being produced [70:31] in multiple ways and in a million different ways. And you, you you can't rely on the [70:38] standard, the conventional qualitative methods or the conventional quantitative assessments, because [70:44] then you actually create a lot of biases in talking about anti-Semitism. Um, as I mentioned [70:51] Before. Uh, just statistical analysis won't help you because there's just, um, [70:58] the sheer amount of indirect speech is just overwhelming. So you need to have models that are [71:05] able to understand to pick up this kind of noise, this kind of, um, ambient anti-Semitism in, in these [71:12] online communities. And for that, you need to have models that are able to do distinguish between [71:17] these different patterns who have enough world knowledge to, uh, to come to a well-informed [71:23] decision. So. That the, the, the, the I wouldn't [71:30] see any opportunity, any way of, of shedding light on these current developments without using [71:36] sophisticated tools that AI, uh, produces. Right. And, and so [71:43] we can only use the same, I guess it would lead us, it would push us in this direction that we [71:49] actually need to use the same tools that are very often misused by [71:56] opponents in order to understand what is driving. Um, society discourse and the future [72:03] of democracy and so on. Um, because it is the the clear separation [72:09] between qualitative, in-depth analysis and quantitative, usually rather superficial [72:15] understanding of contemporary discourse. The bridge between them is AI, right? [72:22] So in depth and representativeness can only come together if we have tools that are able to [72:29] to understand these like this huge diversity and complexity of of human [72:35] communication. If people want to learn more about your work, learn more about all of [72:42] this, where can they? So we have um, there's of [72:49] course, the Decoding Antisemitism website that also shows a range of applications. You can also [72:55] find the Decoding Antisemitism lexicon, uh, on that website. We are now here at NYU [73:02] developing, uh, a new website for the new project called Decoding Hate, where we're looking where we [73:09] will be looking at anti-Semitism, but also into anti-Asian and anti-Black racism and misogyny, [73:14] which is just the start. We will open it up towards other hate ideologies as well. So this [73:20] website is, um, in preparation and will be launched hopefully in two months from now. [73:27] And apart from that, we have a Substack channel, uh, called Decoding Antisemitism. [73:33] And uh, are very, very happy to to get in touch with interested, um, cooperation partners [73:40] who work in education, in law, in, uh, in the political arena in order to share more of our [73:47] insights. Outstanding. Matthias Becker, thank you so much for joining me. Thank you so much. [73:58] Here's what stayed with me. Anti-Semitism, he said, is never really about the Jews. It's about the [74:04] anti-Semites. It's a magnifying glass. You hold up to society, and what you see through it is the [74:11] society itself, its fears, its resentments, who it blames when things go wrong. [74:17] And that's why he reads the comment section closely. It tells you where a country's actually [74:24] headed, and the part I didn't expect after ten years staring into the worst of it. He isn't a [74:31] cynic. He's hopeful because the same technology that lets one person manufacture a wave of [74:37] propaganda is also the first tool we've had. Powerful enough to read the flood back to [74:44] see the coded hate at scale. The virus and the antidote come from the same lab. [74:51] Which one wins is a choice. You can find Matthias's work at the decoding [74:57] anti-Semitism project and on his Substack, Decoding Antisemitism. His team's lexicon — the [75:04] field guide to all of this — has been downloaded more than 350,000 times, and it's free. Links are in [75:10] the description. If this one made you see your own feed differently. Do me a favor. Leave the show a [75:17] rating and a review on Apple or Spotify, wherever you're listening. It's the single biggest thing [75:24] that helps new people find us. And if you're not already subscribe, hit the notification bell and [75:30] follow Honest Reporting on Instagram, X, TikTok and Facebook. I'm Ben Chertoff. This is The Honest Take [75:36] and we'll see you next time.
