Who is Shaping the Cannabis Narrative? Insights from Verified, Regular, and Suspended Tweets

TL;DR

As cannabis legalization sweeps across the U.S., social media has become the primary battleground for public opinion. This study analyzes over 700,000 tweets to reveal a stark divide in how different user "classes" talk about weed. While verified users act as industry news anchors, suspended users are stealthily operating as high-energy sales engines, leaving regular users caught in the middle with the most diverse—yet least positive—discourse.

Problem & Motivation: Beyond the "Average User"

Most public health research on social media treats all accounts as equal. However, a tweet from a verified news outlet or a celebrity carries different weight than a "bot-like" account destined for suspension. The researchers identified a critical gap: we don't know how these different user roles influence the cannabis conversation.

The insight here is that by segregating users by account status, we can see the "agendas" behind the content. Why do some accounts get banned while others thrive? How does this affect what a teenager sees when they search for #420?

Methodology: Qualitative Depth meets Quantitative Scale

The researchers didn't just rely on algorithms. They used a "Mixed-Methods" approach to ensure both statistical power and human intuition:

1. Direct Observation: Manually coding 3,000 tweets to build a thematic framework.
2. LDA Topic Modeling: Using Latent Dirichlet Allocation to discover hidden structures in 724,012 tweets.
3. Sentiment Mining: Using VADER (Valence Aware Dictionary and sEntiment Reasoner) to map the emotional landscape of the discussion.

Fig 1: Determining the optimal number of topics for each user group using coherence scores.

The Core Discovery: A Tale of Three Tiers

The study found that while all users care about Legalization, their motivations vary wildly:

• Verified Users (The Influencers): Focus on the "Macro." They talk about business, policy, and long-term industry growth. They have the highest reach (median 17,326 followers).
• Suspended Users (The "Sales" Engines): These users were significantly more likely to tweet about Cultivation and Advertising. Interestingly, they had the highest sentiment scores—likely because they are "selling" a product or a lifestyle with high-energy, positive language.
• Regular Users (The Public): Their content is the most "human," filled with jokes, daily recreational stories, and diversified opinions.

Fig 2: The intersection of topical interests highlights Legalization as the universal theme.

Why Sentiment Matters

A counter-intuitive finding was that Suspended Users had more positive sentiment (0.0594) than Regular Users (0.0402). In the world of social media moderation, "malicious" doesn't always mean "angry." Often, it means "overly promotional." These accounts use positive emotion to bypass filters or lure users into commercial transactions that violate platform policies.

Fig 3: The distribution of sentiment scores across the dataset shows a predominantly neutral-to-positive lean.

Critical Analysis & Conclusion

This paper provides a roadmap for Public Health 2.0. Instead of shouting into the void, health practitioners should:

• Collaborate with Verified Accounts: Since they have 42x the reach of regular users, they are the most efficient "megaphones" for safety information.
• Monitor Suspended Account Patterns: By understanding the "sales pitch" used by accounts that eventually get banned, educators can warn youth about predatory marketing.

Limitations: The study relies on VADER, which may struggle with the heavy use of slang (e.g., "this kush is fire") common in cannabis culture. Future work using Large Language Models (LLMs) could better decode the nuance of internet "slanguage."

Takeaway: The "who" behind the message is just as important as the message itself. Professionalizing public health outreach means understanding the social hierarchy of the platform.