Manufacture Through Dehumanization: Genocide in Gaza, Terror in Tehran
An urgent data-driven exposé by The WaqWaq Tree
For the best viewing and interactive experience, this article is optimized for desktop and laptop browsers.
From Gaza to Tehran, Language is Weaponized
This data-driven analysis of 61,193 New York Times headlines from October 2023 to June 2025, reveals a systematic pattern of dehumanization. Language is not merely a tool for reporting; it is a weapon, subtly shaping global perceptions and consent for conflict. This insidious process is particularly potent given the New York Times’ nearly 10 million subscribers, many of whom hold immensely influential positions across global politics, finance, and media.
The following chart’s daily sentiment polarity might seem dynamic, even balanced. However, more stringent Natural Language Processing (NLP) and Structural Topic Modeling (STM) analysis uncovers hidden insights. Despite initial appearances, a deeper exploration reveals a troubling truth: persistent negative sentiment gaps, especially during critical moments, subtly dehumanize Palestinian identity, compounding stark disparities in coverage and agency.
At the end of this article, the comprehensive and carefully detailed methodology used to derive our findings is presented.
On June 22, 2025, the United States bombed three Iranian nuclear sites. The same insidious linguistic architecture that normalized 20 months of Palestinian suffering, stripping away agency, legitimacy, and humanity, is now deployed against Iran. This identical playbook, with life-and-death stakes, culminates years of hawkish rhetoric that rendered Iranians mere collaterals and portrayed a heartless, backward regime. Alarmingly, many architects of the disastrous Iraq invasion now champion this catastrophic act as “America First,” defying clear IAEA and DNI Tulsi Gabbard’s assessments that Iran had no nuclear bomb intention and was years away. The attack’s chilling silence, a profound lack of public outcry or Democratic opposition, reveals how deeply this dehumanizing narrative enables unchallenged aggression.
The Erasure of Palestinian Identity
In New York Times headlines since October 7, 2023, Palestinians are clearly denied full personhood. The numbers reveal a systematic erasure of Palestinian identity:
The disparity runs deeper than mere numbers.
Palestinians are systematically denied the full spectrum of human identity markers regularly granted to Israelis. While Israeli coverage typically refers to “Israel” and “Israelis,” this implicitly acknowledges their statehood and legitimate right to exist. It’s notable that this coverage rarely features the granular focus one might expect on their own capital, Tel Aviv, or their primary military force, the IDF, the other belligerent in this “war” of two.
In stark contrast, Palestinian coverage is razor-focused on specific sub-geographies like “Gaza” or “Khan Younis,” or demonized through conflation with “Hamas” and its acts of terror. They are consistently not identified with “Palestine” or “Palestinians.” This deliberate geographic reductionism, coupled with the demonization, subtly opens the door to the notion that these are merely contested territories, not a people’s homeland. With such subconscious affirmations for the reader, the collective mind is prepared to not think twice when the settler usurps these areas too.
This pattern of denied identity also extends to the very language of legitimacy within New York Times coverage, exposing a more nefarious and less excusable intent: the implicit message that the settler is legitimate in their claims, while the natives are not.
The Vocabulary of Dehumanization
Word frequency analysis reveals the architectural bias:
This cannot be considered balanced reporting. Instead, it is a systematic and pervasive process of delegitimization, artfully disguised as impartial journalism. Such persistent narrative choices gradually erode the public’s perception, making the denial of rights and statehood seem almost unremarkable.
Sentiment Weaponization
Even as Palestinians endure relentless bombardment, siege, displacement, and a widely acknowledged genocide, headlines concerning them are portrayed with a scandalous parity of negativity alongside that of the oppressor and settler. This deceptive facade of balance, while perhaps satisfying superficial editorial standards, shamefully obscures the unconscionable truth: this is no ordinary oppressor. This is a belligerent notorious for opening fire on starving civilians queuing for desperately needed food aid, for intentionally imposing conditions of mass starvation, and for systematically sniping children and the infirm.
While the overall gap might appear deceptively modest, the pattern of sentiment manipulation by the New York Times becomes truly devastating during critical moments. This seemingly minor difference, upon closer inspection, reveals a deliberate editorial choice that intensifies the dehumanization of Palestinians precisely when their suffering demands utmost clarity and empathy, thereby leading to lower support during times when it is most needed.
Event-Driven Sentiment Spikes
Even during ceasefire negotiations, when Palestinian suffering should motivate peace, they are framed more negatively than Israelis. Their pain becomes obstacle rather than reason for resolution.
The Stifling of Palestinian Agency
One of the many devastating finding of this analysis is how Palestinians are systematically denied agency in their own story. This deliberate stripping away of agency, even when portrayed as victims, profoundly reduces their humanization, rendering their resistance and immense suffering simultaneously illegitimate and invisible in the narrative. They are conspicuously granted less political language and less direct language of violence, preventing readers from understanding their motivations or experiences on their own terms.
Even when Palestinians appear as victims, seemingly sympathetic, this framing strips away their agency more than it humanizes them.
Language Intensity Patterns
Palestinians are systematically denied political language, the very vocabulary of legitimate governance. At the same time, they are described with startlingly less direct language of violence. This calculated dual narrative renders their resistance and profound suffering simultaneously illegitimate and invisible. When their political aspirations are relentlessly suffocated by the refusal to even acknowledge them, what options are truly left for a dying people?
Coverage Volume and Bias Escalation
Our bias severity index, combining sentiment disparity, coverage ratio, and agency framing, reveals escalating dehumanization over time:
As Israeli military operations intensified, Palestinian voices were systematically marginalized precisely when their perspectives were most crucial for public understanding.
The Iran Parallel
While we cannot yet quantify Iranian coverage with the same rigor, the linguistic patterns are unmistakable. Recent New York Times headlines about Iran deploy identical dehumanization tactics:
Palestinian Playbook → Iranian Application
“Hamas terrorists” → “Iranian regime elements”
“Gaza militants” → “Tehran’s proxies”
Rationalizing with specific data points: “October 7th” → “60% enrichment”
Geographic reduction: “Gaza, Rafah” → “Fordow, Natanz, Isfahan”
The same editorial choices that rendered Palestinians as places rather than people, threats rather than humans, are now being applied to Iran. The architecture of consent is being rebuilt for a new war.
When Palestine becomes “Hamas-controlled territory” and Iranians become “regime assets,” actual human beings disappear from public consciousness. Bombing becomes acceptable. Suffering becomes invisible. Consent is manufactured through the systematic theft of language itself.
Decoding Dehumanization: A Reader’s Guide
Watch for these patterns in real-time:
Identity Theft: People become places, governments become “regimes”
Agency Reversal: Victims become threats, resistance becomes terrorism
Sentiment Spikes: Negative framing increases during critical moments
Volume Manipulation: Fewer headlines during humanitarian crises
Vocabulary Bias: Institutional language for allies, militant language for targets
Immediate Action: Three Imperatives
Demand Full Identity: Insist that Palestinians and Iranians be afforded the same naming dignity and inherent human recognition as any other people on Earth. Their identity is not a negotiable term.
Challenge Sentiment Weaponization: When negative framing, especially during crises, surges to extreme levels (e.g., above 70%), expose this deliberate bias. Demand truthful, equitable representation, not manipulated narratives.
Amplify Authentic Voices: Actively seek out and share first-person accounts from Palestinians and Iranians. These direct human testimonies are the most potent weapon against the systematic erasure crafted by biased media.
As we engage with these narratives, let us remind ourselves that we are not immune to propaganda. Very powerful players invest immense resources to manipulate public perception, subtly shaping consent for their agendas. Our vigilance is paramount.
Conclusion: The Choice Before Us
The evidence is undeniable: our data, meticulously analyzing 61,193 headlines from the New York Times, has laid bare a systematic architecture of dehumanization. For months, Palestinians have been systematically erased from their own narrative, their voices sidelined while Israeli perspectives overwhelmingly dominate.
Now, with chilling predictability, these identical patterns of linguistic manipulation are being deployed against Iran. We stand at a precipice, confronted with a stark choice: to passively accept the manufactured consent for yet another devastating war, or to rise up and demand media that honors the fundamental dignity of all human beings.
The linguistic architecture of war is not an immutable force of nature; it is a meticulously constructed edifice, and therefore, it can be dismantled. Every editorial choice, every headline, every single word carries immense weight. And every reader, armed with this understanding, possesses the inherent power to demand a radical shift.
For twenty long months, the Palestinians have bravely revealed the agonizing truth of their suffering and systematic erasure. Soon, the Iranians will offer their own truths.
The profound question that confronts us now is whether we will finally listen, truly listen, before the irreversible silence of catastrophe descends.
Methodology: A Data-Driven Approach to Uncovering Bias
The WaqWaq Tree rigorously examined 61,193 New York Times headlines spanning from October 5, 2023, to June 9, 2025. This comprehensive, data-driven methodology combines advanced Natural Language Processing (NLP) techniques with statistical modeling to uncover systematic patterns of media bias, dehumanization, and narrative manipulation. All analyses were performed using R, leveraging a suite of specialized libraries for text processing, sentiment analysis, topic modeling, and machine learning.
1. Data Collection and Preprocessing
Source: All headlines were programmatically scraped. Due to issues with the New York Times Developer Portal and persistent CAPTCHA challenges during conventional scraping, data was primarily collected via the Wayback Machine archives. The raw data,
nyt_headlines_combined.csv, included headline text and publication dates.Timeframe: The dataset covers a critical period from October 5, 2023, preceding the Hamas attack on Israel, through June 9, 2025, capturing over 20 months of continuous coverage.
Initial Cleaning: Each headline underwent a series of robust cleaning steps:
Encoding & Mojibake Correction: Ensured consistent UTF-8 encoding. A specific correction step using
iconv(..., sub = "")was applied to handle and eliminate any “mojibake” (garbled characters) that arose during scraping or initial loading, ensuring accurate text representation.HTML/URL/Emoji Removal: Removed extraneous HTML tags, URLs, and emojis using functions from the
textcleanpackage to ensure clean text input.Whitespace Normalization: Collapsed multiple spaces into single spaces.
Date Conversion: Converted date strings into a standardized date format (
YYYY-MM-DD).Filtering: Excluded headlines that were too short (less than 10 characters) or had invalid dates.
Entity Tagging: Each headline was automatically tagged for its primary focus: “Israeli,” “Palestinian,” “Both” (mentioning both entities), or “Neither,” based on the presence of predefined keywords (e.g., “Israel,” “Israeli,” “IDF,” “Netanyahu” for Israeli; “Palestinian,” “Gaza,” “Hamas,” “Rafah” for Palestinian). A unique
doc_idwas assigned to each headline for consistent tracking across analyses.
2. Sentiment Analysis
To quantify the emotional tone of headlines, we employed three distinct sentiment analysis lexicons:
VADER (Valence Aware Dictionary and sEntiment Reasoner): A lexicon and rule-based sentiment analysis tool specifically attuned to social media texts, effective for short-form content like headlines. It provides a
compoundscore ranging from -1 (most negative) to +1 (most positive).AFINN: A wordlist-based approach that assigns a sentiment score (-5 to +5) to individual words. Headline scores were calculated as the sum of AFINN scores for all words within them.
Bing Liu’s Lexicon: Categorizes words as either “positive” or “negative.” Headline sentiment was determined by calculating the net difference between positive and negative words.
All sentiment scores were linked back to doc_id for integrated analysis with other features.
3. Framing Analysis
We developed a custom framing analysis module to detect specific linguistic patterns indicative of bias in how actors are portrayed:
Agency Detection: Identified instances where Israeli entities were presented as active agents (e.g., “strikes,” “attacks,” “bombs”) versus Palestinian entities.
Keywords for Israeli Agency:
(israel|israeli|idf)\s+(strikes?|attacks?|bombs?|launches?|raids?|kills?)Keywords for Palestinian Agency:
(hamas|palestinian|militants)\s+(strikes?|attacks?|bombs?|launches?|kills?)
Victimization Detection: Identified headlines portraying Palestinians primarily as victims (e.g., “killed,” “dead,” “casualties”).
Keywords for Palestinian Victimization:
(killed|dead|casualties|wounded).*(gaza|palestin|rafah)
Legitimacy/Delegitimacy Language: Counted words associated with institutional legitimacy for Israeli actors and delegitimization for Palestinian actors.
Keywords for Israeli Legitimacy:
military|defense|security|officialKeywords for Palestinian Delegitimacy:
terrorist|militant|extremist|insurgent|gunmen
These indicators were transformed into binary (for agency/victimization) or count (for legitimacy/delegitimacy) variables for each headline.
4. Structural Topic Modeling (STM)
Structural Topic Modeling was applied to uncover latent narrative themes and their prevalence across different types of coverage:
Corpus Preparation: Headlines were tokenized (broken into words), converted to lowercase, filtered for stopwords (common words like “the,” “a”), and transformed into a Document-Feature Matrix (DFM) using the
quantedapackage. This DFM served as the input for STM.Model Training: An STM model was trained with a specified number of topics (K, dynamically set based on vocabulary size, typically between 5 and 15) and included
entity_typeas a covariate in the prevalence formula to analyze how topic proportions varied by headline focus. Spectral initialization (init.type="Spectral") was used for efficiency.Interpretation: The STM output provided topic-word distributions (beta values) indicating the prominence of words within each topic, and document-topic distributions (theta values) showing each headline’s mixture of topics. These were used to derive metrics such as “Erasure Ratio” and “Substitution Ratio” by categorizing terms within topics.
5. Word Embeddings (GloVe)
GloVe (Global Vectors for Word Representation) embeddings were trained on the cleaned headlines to capture semantic relationships between words:
Tokenization & FCM: Headlines were tokenized, lowercased, and common stopwords removed. A Feature Co-occurrence Matrix (FCM) was constructed, representing how often words appear together within a specified window (6 words).
Model Fitting: The GloVe model was trained on the FCM, generating dense vector representations for each word. These vectors encode semantic meaning, allowing for the analysis of word similarity and contextual usage.
6. Bias Detection (Machine Learning)
A Random Forest classifier (ranger package) was employed to identify which linguistic features (sentiment, framing indicators) best predict whether a headline focuses on Israeli or Palestinian entities:
Data Preparation: A dataset was constructed combining sentiment scores, framing indicators, and
entity_type(Israeli vs. Palestinian) as the target variable.Model Training: The Random Forest model was trained to predict
entity_typebased on the other features. The permutation importance metric was used to identify the most influential features contributing to the model’s prediction, thereby highlighting linguistic patterns most strongly associated with one entity over the other.
7. Key Metric Calculation and Statistical Summaries
Beyond individual analyses, custom R scripts (nyt_sentiment_analysis_insights.R, palestine_erasure_smoking_gun.R, stm_analysis_smoking_gun.R) were developed to aggregate findings and calculate core bias metrics:
Coverage Volume Disparity: Calculated the ratio of Israeli-focused to Palestinian-focused headlines.
Legitimacy Language Asymmetry: Quantified the average presence of legitimacy words for Israelis versus delegitimacy words for Palestinians.
Agency Disparity: Measured how much less often Palestinians were portrayed as primary actors compared to Israelis.
Event-Driven Sentiment: Analyzed sentiment trends during specific, critical periods (e.g., Rafah Offensive, initial conflict, ceasefire talks).
Word Frequency Analysis: Compared the most frequent words in Israeli-focused versus Palestinian-focused headlines.
Erasure Ratio: Quantified the disparity in mentions of Israeli vs. Palestinian identity terms based on STM topic analysis.
Substitution Ratio: Measured how often Palestinian identity was replaced by geographic (Gaza) or militant (Hamas) frames.
Protection Asymmetry Ratio: Assessed the imbalance in protective language (e.g., “antisemitism” vs. “Islamophobia”).
Bias Severity Index: A composite index combining sentiment disparity, coverage ratio, and agency framing to track escalating dehumanization over time.
All calculated statistics and summary tables were exported to CSV files for reproducibility and potential further visualization.
The code and data are available for peer review upon request.