The Structure of Participation and Attention in Arabic-Language Hezbollah Discourse on X

1. Problem Statement

The study addresses a critical gap in understanding the dynamics of political discourse on social media platforms, specifically regarding the disparity between participation (who posts) and attention (who gets seen/engaged). While platforms like X (formerly Twitter) appear to offer open, democratic spaces for political discussion, there is a need to quantify whether audience attention is distributed evenly or concentrated among a specific elite. The research focuses on Arabic-language discourse surrounding Hezbollah, a major political and military actor in the Middle East, to determine:

• How engagement is distributed across the user base.
• Whether specific account types (e.g., institutional media vs. ordinary users) play a disproportionate role in capturing attention.
• The structural dynamics of influence in a high-stakes geopolitical context.

2. Methodology

Data Collection

• Source: X (formerly Twitter) via the Apify cloud platform.
• Timeframe: March 1–8, 2026.
• Search Strategy: Arabic-language queries containing "حزب هللا" (Hezbollah) and spelling variations.
• Filters:
  • lang:ar to restrict to Arabic content.
  • -filter:retweets to exclude retweets, ensuring the dataset captures original content generation rather than amplification.
  • Exclusion of automated AI responses (specifically the "Grok" account).
• Dataset Size: 15,767 unique tweets from 8,148 unique users.

User Classification

Users were categorized into two operational groups based on metadata (username, display name, bio) using a keyword-based approach:

1. Media-labeled accounts: Contain keywords indicating journalism or media activity (e.g., "channel," "newspaper," "broadcaster," "agency"). This category was refined with manual overrides for clear institutional outlets and parody accounts.
2. Non-media users: All other accounts (individuals, activists, commentators).

• Resulting Split: 844 Media users (10.4%) vs. 7,304 Non-media users (89.6%).

Engagement Metric

• Definition: Total engagement = Likes + Reposts + Replies.
• Exclusion: Quote tweets were excluded to avoid attribution ambiguity (as the interaction happens on the quoting tweet, not the original).
• Analysis: Engagement was aggregated at both the tweet level (average per post) and user level (total cumulative engagement) to measure concentration.

3. Key Contributions

• Quantification of Attention Inequality: Provides empirical evidence of extreme engagement concentration in Arabic political discourse, demonstrating that a "participatory" conversation does not equate to a "distributed" attention economy.
• Media vs. Non-Media Dynamics: Offers a granular comparison of how institutional media accounts perform against ordinary users in terms of both volume of production and efficiency of engagement.
• Operational Classification: Introduces a reproducible, keyword-based method for classifying media actors in Arabic social media discourse, acknowledging the nuance between institutional outlets and individual "media-labeled" personalities.

4. Key Results

A. Extreme Concentration of Attention

The distribution of engagement follows a highly skewed power law:

• Top 1% of users (81 accounts): Capture 61.5% of all total engagement.
• Top 5% of users: Capture 90.6% of total engagement.
• Top 10% of users: Capture 96.2% of total engagement.
• Implication: While thousands of users participate, the vast majority of audience attention is funneled to a tiny elite.

B. Participation vs. Production

• Volume: Non-media users dominate content production, generating 79.9% of all tweets (12,603 tweets) compared to 20.1% from media accounts.
• User Base: Non-media users constitute 89.6% of the participant pool.

C. The "Media Advantage"

Despite producing fewer posts, media-labeled accounts are significantly more efficient at generating engagement:

• Average Engagement per Tweet:
  • Media accounts: 41.32 interactions.
  • Non-media accounts: 30.84 interactions.
  • Result: Media posts generate roughly 34% more engagement per tweet on average.
• Total Engagement Share: Despite the efficiency gap, non-media users still account for 74.8% of total engagement due to their sheer volume of posts, while media accounts account for 25.2%.

D. Composition of the Elite (Top 1%)

Among the 81 users capturing 61.5% of engagement:

• 29.6% are media-labeled accounts (significantly overrepresented compared to their 10.4% share of the total user base).
• 70.4% are non-media users (including political commentators, activists, and high-following parody accounts).
• Follower Count: The majority of these top accounts have >10,000 followers, with >50% exceeding 100,000 followers.

5. Significance and Conclusion

The study concludes that Arabic-language discourse on X regarding Hezbollah exhibits a structural duality:

1. Broad Participation: The conversation is driven by a massive base of ordinary users who generate the majority of content.
2. Concentrated Visibility: Audience attention is overwhelmingly monopolized by a small minority of highly visible accounts, a mix of institutional media and influential non-media personalities.

Significance:

• Political Communication: The findings suggest that while social media democratizes speech (allowing many to post), it does not democratize attention. Political narratives are still largely shaped by a small, highly visible elite.
• Media Influence: Institutional media retains a structural advantage in visibility, but they share the "attention economy" with influential individual actors and satirical accounts, indicating a complex ecosystem where traditional media authority competes with digital influencers.
• Methodological Insight: The study highlights the importance of distinguishing between posting activity and engagement metrics when analyzing political discourse, as high participation rates can mask extreme inequality in influence.

Limitations: The "media" classification is an operational proxy based on self-identification in metadata rather than a strict institutional audit, and the data is restricted to a specific one-week window during a specific geopolitical period.