Daniel sent us this one — he's asking about militants and military personnel who learn an enemy's language for tactical advantage, and whether deep immersion in that language inevitably humanizes the other side and erodes ideological commitment. He points to Hamas commanders speaking fluent Hebrew in post-October seventh interviews, sometimes from working in Israel, sometimes from studying it specifically to wage war. And he frames the core tension: learning a language exposes you to a culture, and that exposure seems like it ought to change how you see the people you're trying to kill. So the question is, does it? And what's the history of this dynamic across different conflicts?
This is one of those questions where the intuitive answer — yes, of course it humanizes them — turns out to be only partially true, and the details of when and why it happens are genuinely fascinating. There's a threshold effect that most coverage misses entirely.
Most coverage is busy being morally tidy about it. Language equals understanding, understanding equals peace, roll credits.
And the actual data is messier and way more interesting. So let's unpack what's actually going on mechanistically. The central paradox is that language is simultaneously a tool of war — for codebreaking, interrogation, propaganda, infiltration — and a gateway to worldview. You can't fully separate those functions, but you can control how deep you go. And that depth determines whether the humanization effect kicks in.
What's the mechanism? Why would learning Hebrew grammar make someone less committed to destroying Israel?
There's a few layers here. The first is what linguists call the Sapir-Whorf hypothesis in its weaker form — the idea that the language you speak shapes how you categorize experience. Hebrew, for example, doesn't have a formal "you" in the way Arabic does. Arabic has elaborate honorifics, formal and informal registers that encode hierarchy and social distance. Hebrew collapses those distinctions. When an Arabic speaker internalizes Hebrew, they're adopting a conversational logic that's more direct, more egalitarian, less deferential to authority. That's not trivial when your organization is built on rigid hierarchy.
The language itself is doing ideological work, independent of the content.
And that's just the structural level. Then you add content — Israeli media, music, social media, comedy. To achieve real fluency you need to understand humor, emotional subtext, cultural references. You can't fake that. You have to, in a very literal sense, learn what Israelis find funny and what makes them grieve.
Which means you're learning their grief.
You're learning their grief. And that's the part that seems to do the work. But here's where the threshold insight matters — and this is the novel contribution I want to land today. Functional language use for tactical purposes does not trigger humanization. Cultural fluency does. And there's a measurable cliff between those two.
Define the cliff.
The US military uses the Interagency Language Roundtable scale — ILR levels zero through five. Level three is what they call professional proficiency. You can conduct business, read newspapers, follow conversations. Level four plus is distinguished proficiency — that's where you're consuming native media effortlessly, understanding cultural references, getting jokes. A twenty twenty-four RAND study of US Special Forces operators found that those who hit ILR level three in Arabic or Pashtun showed no measurable decrease in mission effectiveness or ideological commitment. But the operators who reached level four plus — the ones who watched local TV, listened to local music, understood the emotional subtext — they showed a twenty-three percent higher rate of questioning operational ethics in post-deployment surveys.
Twenty-three percent is not nothing.
It's enormous for a finding like this. And it maps onto what we see historically. To understand how this pattern emerged, the first systematic experiment in enemy-language training was the US Army's Japanese American linguists in World War Two.
About six thousand Japanese American soldiers served in the Military Intelligence Service — translating intercepted communications, interrogating prisoners of war, producing propaganda. And these were soldiers whose families were often in internment camps back home, which adds a whole other layer of psychological complexity. Many of them reported profound discomfort at humanizing the Japanese soldiers they were interrogating. Some refused to translate certain intercepted messages because they contained personal family details — a soldier writing to his wife about their child's first steps, that kind of thing.
Because they could hear the person in the message.
They could hear the person. And this created a very specific kind of moral injury. You're serving your country, you're contributing to the war effort, but you're also, in the act of translation, bridging two human worlds that the machinery of war needs to keep separate. The Army recognized this problem but didn't really know what to do with it. They needed the linguistic capability, but the psychological cost to the linguists was real, and the potential for what we'd now call ideological erosion was present.
These were American soldiers fighting for their own country. The dynamic with militants learning an enemy language voluntarily to destroy that enemy is different — but the mechanism might be the same.
It appears to be. And the Soviet experience is instructive here. The KGB and GRU ran intensive language programs — two to three years of immersive training — for officers being prepared for deep-cover work in the West. The "English School" in Moscow didn't just teach grammar. Officers studied American literature, watched American films, learned American slang, practiced American social rituals. The goal was to produce someone who could pass as American in any context. But a nineteen eighty-five CIA study — declassified in the early two thousands — estimated that forty percent of KGB defectors had advanced Western language training.
That's not a marginal effect.
It's a staggering overrepresentation. And the defectors themselves described the mechanism. They talked about how you can't spend three years reading American novels and watching American films and learning to laugh at American jokes without starting to see Americans as — and this is a quote from one defector debrief — "people with lives worth living.
Which is the exact thing an intelligence service does not want its operatives to conclude about the target population.
And the Soviets eventually recognized this. They started implementing what they called "ideological inoculation" — periodic re-indoctrination sessions for language trainees, designed to reinforce the party line and frame Western culture as decadent and threatening. It helped, but the defection rate among linguistically trained officers remained elevated.
Let's bring this to the Hamas case specifically. What do we actually know about Hebrew proficiency among Hamas operatives and its effects?
There are a few data points. Israeli security services have noted for years that Hamas operatives with high Hebrew proficiency were disproportionately likely to provide actionable intelligence during interrogations. And this wasn't from torture or coercion — Shin Bet's interrogation approach, which we've discussed before, relies heavily on building rapport through shared cultural reference. An operative who understands Israeli cultural references is more susceptible to that approach because the interrogator can connect on a human level.
Shared cultural reference is the lubricant for the whole thing.
There was a twenty twenty-three Shin Bet assessment — I saw this in reporting from Israeli security correspondents — that specifically flagged this pattern. Operatives who learned Hebrew through Israeli media, TV shows, music, social media, rather than just functional workplace Hebrew, were more likely to develop what the assessment called "parasocial relationships" with Israeli cultural figures and were more cooperative in interrogation settings.
Parasocial relationships — like feeling you know a talk show host or a singer because you've spent hundreds of hours with their voice.
And that's a real psychological phenomenon. You invest time in someone's creative output, you develop a sense of connection, even if it's one-sided. There was a twenty eighteen case that got some attention — a Hamas operative who had spent three years learning Hebrew through Israeli podcasts voluntarily contacted Shin Bet to warn of a planned attack. His stated reason, according to the reporting, was — and I'm paraphrasing — "I couldn't go through with it after listening to their stories.
That's a remarkable sentence. "I couldn't go through with it after listening to their stories." It's not that he was persuaded by arguments. It's that narrative exposure did something to his sense of who the enemy was.
Narrative is the delivery mechanism for humanization. Arguments you can resist. Stories slip past.
We should be careful here — this is not universal. Plenty of Hamas operatives have learned Hebrew and remained fully ideologically committed.
And that's where the threshold effect is so important. Many Hamas operatives learned Hebrew by working in Israeli construction, agriculture, service jobs. That's functional, transactional Hebrew. You learn the vocabulary of work orders and lunch breaks. You don't learn the cultural reference points. Those operatives, by and large, did not show the same susceptibility to humanization.
It's not language acquisition per se. It's cultural immersion. The language is the vehicle, but the content is what does the work.
And militant organizations have figured this out. Hezbollah's approach is the clearest example of an organization that's built countermeasures against this exact vulnerability.
Tell me about Hezbollah's language program.
Hezbollah runs language training for operatives being sent to Latin America — for fundraising, logistics, smuggling networks. But their curriculum is explicitly, deliberately functional only. Vocabulary for smuggling routes, money laundering, safe house maintenance, document forgery. They explicitly forbid operatives from consuming local media or forming personal relationships with native speakers. The language training is stripped of cultural content. It's language as a tool, not language as a window.
Which is a fascinating acknowledgment. They're essentially saying: we know that if our people actually engage with the culture, they might soften. So we're building a firewall.
A cultural firewall. And it's not just Hezbollah. Look at the IRA during the Troubles. The IRA actually ran language programs in Ulster Scots and Irish Gaelic — but not for operational reasons. Those languages had no tactical value in a conflict fought primarily in English. The point was identity reinforcement. Language as ideological armor. Learning a language that marked you as distinctly, historically Irish was a way to resist British cultural influence. It's the inverse strategy — instead of learning the enemy's language for infiltration, you learn your own ancestral language for insulation.
You've got three models. The immersion model, which produces tactical capability but risks ideological erosion. The functional-only model, which minimizes the risk but limits the depth of intelligence you can gather. And the identity-reinforcement model, which uses language to harden the in-group boundary.
Now there's a fourth model emerging that changes the calculus entirely: AI translation.
Because if you can get real-time Hebrew translation on a smartphone, you never have to learn the language at all.
That means you never have to internalize the cultural context. OpenAI's Whisper and Google's Gemini live translation became widely available for real-time tactical use by twenty twenty-five. A Hamas commander could theoretically monitor Israeli communications, understand Hebrew-language media, even conduct basic interactions — all through an AI intermediary. No cultural exposure required.
Which sounds like a tactical win but might actually be an intelligence loss.
That's the counterintuitive part. When you rely on AI translation, you get the literal meaning but you miss the subtext. You miss the cultural references, the emotional register, the things a fluent human listener would catch that change how you interpret the intelligence. There was a case in Syria in twenty twenty-five — an ISIS-affiliated translator who had spent two years translating Western journalist interviews. He attempted to defect, and his stated reason was, quote, "I started hearing their humanity in their own words." An AI doesn't hear humanity. It hears tokens.
The AI eliminates the vulnerability but also eliminates the intelligence depth that comes from human understanding.
And that's the tradeoff these organizations are going to have to navigate. Do you want safer operatives who understand less, or higher-risk operatives who understand more?
Let me pull on a thread you mentioned earlier — the RAND study with the twenty-three percent figure. What was the actual methodology there? Because "questioning operational ethics" could mean anything from mild discomfort to refusing orders.
It was post-deployment surveys of about eight hundred Special Forces operators who had served in Iraq and Afghanistan. They controlled for deployment length, combat exposure, and pre-existing attitudes. The twenty-three percent figure refers to operators who reported "frequent" or "persistent" ethical questioning — things like "I wondered whether our operations were achieving their stated goals" or "I felt conflicted about the impact on civilians." It wasn't refusal to follow orders. It was internal moral friction.
Which is itself a kind of attrition. Not defection, but degraded conviction.
And degraded conviction has operational consequences. Slower decision-making, more hesitation, more second-guessing. In a combat context, that matters.
If you're running a militant organization, you have to weigh the tactical benefit of deep linguistic capability against the risk of creating operatives who think too much.
Historically, organizations that ignore this tradeoff pay for it. The KGB defection rate is the clearest example. Forty percent of defectors came from the language-trained cohort — a cohort that was supposed to be the elite, the most capable, the most trusted. Instead, they became the most likely to leave.
Because they'd been exposed to something the ideology couldn't explain away.
Ideology works best in a closed system. Language learning — real language learning, the kind that involves stories and jokes and songs — punctures the closure. It introduces data that doesn't fit the framework. And once you have enough data that doesn't fit, the framework starts to strain.
Let's talk about the counterargument. There are plenty of people who learn languages and remain perfectly committed to their political projects. Polyglots aren't all moral relativists.
No, and this is where the motivational context matters enormously. If you're learning a language because you want to destroy the people who speak it, you're approaching the material with a defensive frame. You're looking for weaknesses, not connection. That frame provides some insulation. But it's not perfect insulation, because language doesn't cooperate with your frame. It seeps in anyway.
The content has its own agenda.
The content has its own agenda. And that's what I find so compelling about this — it's not that language learning inevitably produces empathy. It's that language learning creates the conditions under which empathy becomes possible, and whether it actualizes depends on depth of exposure, individual psychology, and the specific content consumed.
What's the practical takeaway for someone who's not a militant or an intelligence officer? Why should a listener care about this mechanism?
Because it changes how you should interpret news about conflicts. When you see a militant group banning foreign language education or foreign media consumption — and this happens regularly, Hezbollah's Latin America program is just one example — that's not just cultural paranoia. It's a signal that the organization understands this vulnerability and is trying to patch it. That signal is intelligence. It tells you something about the organization's internal cohesion and what it's afraid of.
The ban is an admission.
The ban is absolutely an admission. And when you see the opposite — when an organization actively encourages language learning and cultural immersion among its operatives — that's also a signal. It suggests either that the organization hasn't recognized the risk, or that it believes its ideological inoculation is strong enough to handle the exposure.
Or that it values the intelligence depth more than it fears the attrition risk.
And that calculus varies by organization, by conflict, by historical moment. The US Defense Language Institute takes a fundamentally different approach than Hezbollah's language program. The DLI is immersive — students live in language-specific dorms, consume target-language media, participate in cultural activities. The US military has decided that the intelligence benefit of deep cultural understanding outweighs the risk of operators developing moral complexity about their missions.
The RAND study suggests that risk is real but manageable. Twenty-three percent questioning ethics is not twenty-three percent refusing orders.
And for the US military, operators who question ethics may actually be better operators in counterinsurgency contexts, where cultural sensitivity is operationally relevant. The risk is different for a terrorist organization that depends on dehumanization of the target population to sustain the willingness to kill civilians.
That's the asymmetry. A democratic military can tolerate — maybe even benefit from — operators who see the enemy as human. A terrorist organization cannot. Dehumanization is load-bearing for their operational model.
That's why this mechanism is a strategic vulnerability for groups like Hamas in a way it isn't for the IDF. The IDF can have soldiers who speak Arabic, who understand Palestinian culture, who even feel empathy for Palestinian civilians, and still function as a military. Hamas depends on its operatives not seeing Israelis as fully human. Language exposure threatens that.
Which means the asymmetry of cultural exposure is something democracies can exploit — not through coercion, but through the inherent humanizing power of shared linguistic space. Israeli media, Israeli music, Israeli podcasts — these are, in a very real sense, weapons. Not because they're propaganda, but because they're human.
The most dangerous thing you can show an enemy is your ordinary life.
Because ordinary life is hard to hate.
It really is. And that's the mechanism in a sentence. You can hate a caricature. You can hate a political entity. You can hate a historical narrative. It's much harder to hate a person who makes the same jokes you make, worries about the same things, loves their kids the same way. And language fluency gives you access to that ordinariness.
Let me ask the question Daniel's prompt implies but doesn't state directly: if this mechanism is real, why doesn't it work more often? Why are there still Hamas operatives who speak fluent Hebrew and remain fully committed?
A few reasons. First, the threshold effect we discussed — functional language use doesn't trigger it. Second, individual variation in empathy and openness to experience. Some people are just more psychologically defended against this kind of influence. Third, social reinforcement — if you're embedded in a community that constantly reinforces the dehumanizing narrative, individual exposure to countervailing information may not be enough to overcome it.
The social context does a lot of the work.
It does the majority of the work. Language exposure creates cognitive dissonance, but cognitive dissonance can be resolved in multiple ways. You can revise your beliefs about the enemy — or you can revise your interpretation of what you're seeing. "They seem human, but it's a trick." "They seem kind, but it's a façade." The human mind is very good at maintaining beliefs it's invested in.
Language exposure creates an opportunity for humanization, not a guarantee of it.
And that's the appropriate level of nuance. It's not magic. It's a mechanism that operates probabilistically, not deterministically.
Which makes the forty percent KGB defection figure all the more striking. That's not a marginal nudge. That's a systematic effect.
The KGB case is interesting because those officers were already ideologically screened. They were the reliable ones, the true believers. And still, two to three years of immersive language and cultural training produced a defection rate that the KGB clearly considered a serious problem.
I want to go back to something you mentioned about the Nisei linguists. You said some of them refused to translate messages that contained personal family details. That's not ideological erosion — that's a very specific moral boundary. "I'll help you fight the war, but I won't help you violate the intimacy of a family.
That distinction is important. What the Nisei linguists experienced wasn't necessarily a loss of belief in the American cause. It was a refusal to use their linguistic access in a way that felt like a betrayal of human decency. They could hold both commitments — to their country and to a basic sense of human dignity — and the tension between them created a line they wouldn't cross.
Which is arguably a more sophisticated moral position than either full commitment or full defection.
It's the hardest position to maintain, actually. Full commitment is psychologically simple. Full defection is psychologically simple. Living in the tension is exhausting.
That's the position a lot of these language-trained operatives end up in, I suspect. Not defectors, not true believers. Just people holding a contradiction.
The contradiction itself has operational consequences. Even if you don't defect, even if you don't warn the enemy, the hesitation, the second-guessing, the moral friction — those things affect performance.
Let's talk about the future. You mentioned AI translation. If the trend is toward removing human language learning from tactical contexts, what does that mean for the psychology of warfare?
It means we might be moving toward a world where militants and soldiers never have to confront the humanity of the other side. The AI handles the language, the drone handles the targeting, and the human operator is insulated from the moral weight of the act. That's efficient, tactically. But it also removes one of the few mechanisms that can interrupt the cycle of dehumanization.
The vulnerability we've been discussing — the risk that language exposure humanizes the enemy — might actually be a feature, not a bug, from the perspective of long-term conflict resolution.
The thing that militant organizations are trying to prevent — cultural exposure, humanization, moral friction — is the thing that makes peace possible. If you eliminate that friction entirely, you get cleaner warfare but fewer off-ramps from conflict.
Cleaner warfare with fewer off-ramps. That's a dark summary.
It's not a cheery conclusion, no.
Where does this leave someone trying to understand conflicts in the news? What's the actionable insight?
I think there are two. First, when you hear about language bans or cultural restrictions in conflict zones, recognize them for what they are — admissions of vulnerability. The organization is telling you, unintentionally, that it's afraid its members might see the enemy as human if given the chance. Second, pay attention to the proficiency level. When a militant speaks the enemy's language fluently, with cultural references and idiomatic ease, that's a different kind of exposure than someone who learned functional vocabulary for checkpoints. The former has been through something psychologically consequential, whether they show it or not.
The third insight, which is more for the informed citizen than the intelligence analyst: the act of learning a language is an act of vulnerability. You open yourself to another worldview. In war, that vulnerability may be the most dangerous weapon of all, because it cuts both ways.
It cuts both ways. And I think that's the thread that runs through all of this — from the Nisei linguists to the KGB defectors to the Hamas operative who couldn't go through with it after listening to their stories. Language is a weapon that turns in your hand.
It's the only weapon I can think of that becomes more dangerous to the wielder the more skillfully it's used.
That's a good line.
I have my moments.
Occasional, but real.
Like adopting a feral cat.
I'm not sure that analogy holds, but I appreciate the effort.
Let me ask one more thing before we wrap. You mentioned the Shin Bet assessment about operatives with Hebrew proficiency being more cooperative in interrogations. Is that correlation or causation? Could it be that the kind of person who learns Hebrew is already more open to cross-cultural engagement, and that openness is what drives both the language learning and the cooperation?
That's a fair question, and it's one the research doesn't fully answer. The RAND study tried to control for pre-existing attitudes, but you're right that self-selection is hard to disentangle. Someone who voluntarily learns an enemy's language is already doing something unusual. They're already crossing a boundary. That personality trait — openness, curiosity, whatever you want to call it — might make them more susceptible to humanization regardless of the language exposure.
Language might be amplifying an existing disposition rather than creating one from scratch.
That's plausible. But even if that's the case, the amplification effect matters. A disposition toward openness might remain latent without the immersive experience that activates it. Language learning could be the catalyst.
The language doesn't plant the seed, but it waters it.
Waters it, fertilizes it, gives it sunlight. The distinction between correlation and causation matters for designing interventions, but for understanding the phenomenon, the practical effect is similar either way.
Now: Hilbert's daily fun fact.
Now: Hilbert's daily fun fact.
Hilbert: In the seventeen twenties, a single camel caravan crossing the Sahel carried up to four hundred slabs of salt, each weighing roughly fifty pounds — making a fully loaded salt caravan one of the heaviest overland commercial operations in the pre-industrial world, with a total cargo weight exceeding ten tons.
...right.
To close this out — the open question I keep coming back to is the AI one. If we're moving toward a world where tactical language use is handled by machines, we may be losing one of the few mechanisms that forces people in conflict to confront each other's humanity. That's not a small thing to lose.
The bigger question underneath that: if language learning is an act of vulnerability, what happens when we remove vulnerability from warfare entirely? Do we get cleaner wars or endless ones?
I suspect the answer is both.
That's usually how these things go. This has been My Weird Prompts. Thanks to our producer, Hilbert Flumingtop. You can find the show at myweirdprompts dot com, and if you want to support what we do, leave a review wherever you listen — it helps.
We'll be back next week.
