The Enemy from Within: A Study of Political Delegitimization Discourse in Israeli Political Speech
This study presents the first large-scale computational analysis of political delegitimization discourse in Israel, utilizing a manually annotated Hebrew corpus and a novel two-stage AI pipeline to reveal a significant rise in delegitimizing rhetoric over three decades, particularly on social media, among male and right-leaning politicians, and during major political events.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine a democracy as a giant, noisy town hall meeting where everyone is supposed to argue about rules, budgets, and policies. For the meeting to work, everyone has to agree on one fundamental rule: "We are all allowed to be here, even if we hate each other's ideas."
This paper is about what happens when someone stops arguing about the ideas and starts trying to kick people out of the building entirely. The authors call this "Political Delegitimization Discourse" (PDD).
Here is the breakdown of their study, translated into everyday language:
1. The Core Concept: "You Don't Belong Here"
Usually, when politicians fight, they say, "Your plan for the roads is bad." That's a policy debate.
PDD is different. It's when a politician says, "You are not a real Israeli," "You are a traitor," or "Your group is an enemy of the state."
- The Analogy: Imagine a soccer game.
- Normal Debate: "That referee call was wrong!"
- PDD: "You aren't allowed to play soccer; you don't belong on this field, and your team is trying to burn the stadium down."
- The goal isn't to win the game; it's to convince everyone that the other team doesn't have the right to exist in the game.
2. The "Detective" Project
The authors (Naama and Guy) wanted to measure how often this happens in Israel. Since there are thousands of speeches and posts, they couldn't read them all by hand. So, they built a digital detective (an AI).
- The Training: They gathered over 10,000 sentences from three places:
- The Knesset (Israeli Parliament) speeches.
- Facebook posts by politicians.
- News articles.
- The Human Work: They hired humans to read these sentences and label them. They taught the AI to spot the difference between "rude" (incivility) and "existential denial" (delegitimization).
- The Result: They found that about 17% of the sentences they looked at were actually trying to delegitimize someone.
3. What the AI Discovered
Once the AI was trained, they let it scan years of data to answer five big questions:
A. Is it getting worse over time? (Yes)
- The Trend: In the 1990s, this kind of talk was rare. By the 2020s, it has skyrocketed.
- The Analogy: It's like a town hall that used to be mostly polite, but over 30 years, the shouting has turned into people trying to lock the doors on each other. The study shows a sharp jump in this behavior, especially after 2020.
B. Is it worse on Facebook or in Parliament? (Facebook)
- The Difference: The amount of delegitimization is similar on both platforms, but the flavor is different.
- The Analogy: Parliament is like a formal dinner where people might whisper insults. Facebook is like a mosh pit where the insults are louder, more personal, and more aggressive. The AI found that on Facebook, politicians are more likely to call opponents "enemies," use swear words, and attack specific people rather than just groups.
C. Do men or women do it more? (Men)
- The Finding: Male politicians use this "you don't belong" language slightly more often than female politicians.
- The Analogy: If the town hall were a boxing ring, the men were throwing more "disqualification" punches than the women.
D. Do the political sides (Left vs. Right) do it differently? (It depends on the time)
- The Long View: Over the last 30 years, the Right has generally used this rhetoric more.
- The Short View: In recent years (2018–2021), the Left actually spiked and used it more than the Right.
- The Twist: The Right tends to delegitimize entire groups (e.g., "The Left is evil"), while the Left tends to focus more on specific individuals.
E. Does it happen more during elections? (Yes)
- The Pattern: This behavior is like a weather vane. It spins wildly during election campaigns (when politicians are fighting to win) and calms down when a government is formed and they need to work together.
- The Analogy: When the election is on, everyone is screaming "Get out!" Once the election is over and they have to share a house (form a coalition), they stop trying to kick each other out, at least for a while.
4. Why This Matters
The authors argue that previous computer programs were good at spotting "bad words" (swearing) or "hate speech" (racism), but they missed this specific, subtle type of attack.
- The Gap: You can say something very polite, like "This person is a danger to our democracy," without using a single swear word. Old computers missed this. The new AI they built can spot the intent to deny someone's right to participate, even if the words sound "civil."
Summary
This paper is a warning and a tool. It warns us that in Israeli politics, the habit of saying "You don't belong here" is becoming much more common, especially on social media and during elections. It also provides a new "radar" (the AI model) that can automatically track this dangerous trend in the future, helping us understand when a democracy is starting to break down from the inside.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.