NexusSIM Research Brief

How Online Networks Transformed Climate Skepticism

Linking a conspiratorial counterpublic to a foundational event.

J. Connor Balch · Just Horizons Alliance · 2026

The Problem

Climate change is one of the most polarized topics in American public life, and has become deeply entwined with cultural identity and partisan boundaries. Conventional explanations of climate skepticism often portray it as a failure of science communication, but this fails to account for the deeper emotional and social dynamics that fuel the adoption and dissemination of misinformation, and leaves many open questions. Why does climate skepticism persist even as scientific evidence accumulates? Why do online communities of climate denial seem to grow rather than shrink in the age of information abundance?

This research investigates climate skepticism not as a simple failure to understand science, but as a culturally embedded worldview with its own internal logic, emotional dynamics, and networked infrastructure. Drawing on cognitive anthropology and computational social science, it maps how anti-environmentalism evolved from a coherent elite conservative ideological project into a diffuse, conspiratorial digital movement.

Core argument: Climate skepticism functions as a ritualized counterpublic—a self-reinforcing community bound together by shared affect, identity performances, and conspiratorial narratives, sustained through social media's capacity for rapid collective mobilization.

Data & Methods

This work draws on a large, multi-source corpus spanning over a decade of climate discourse across media contexts that differ in their institutional constraints, audiences, and norms:

4
Text Corpora
12 yrs
Twitter Data (2008–2020)
4
Computational Methods

The textual corpora include conservative think tank publications (e.g., The Heartland Institute), cable news transcripts (Fox News, CNN, MSNBC), climate skeptical blogs (e.g., Watts Up With That), and millions of climate-related tweets. Four complementary computational methods are applied across these data:

Topic Modeling

BERTopic models identify thematic clusters in each corpus, revealing how climate change is framed differently across institutional contexts.

Word Embeddings

Word2Vec models trained per-source capture the semantic neighborhoods of climate change—what concepts are nearby in each discourse.

Network Analysis

Co-retweet networks trace how climate skeptic communities form, polarize, and persist on Twitter from 2008 to 2020.

Claim Classification

A fine-tuned RoBERTa model automatically detects and categorizes specific types of climate denial in tweets.

The analysis integrates four computational approaches, each capturing a different dimension of the discourse:

BERTopic uses embedding models to cluster documents by semantic similarity, surfacing the thematic structure of each corpus by identifying what topics dominate and how they differ across sources. Word2Vec builds up implicit associations between word or phrase-level concepts from patterns of co-occurrence, mapping the semantic neighborhood of key concepts like “climate change” in each text dataset. This algorithmic approach operationalizes cultural schema theory: a theory in cognitive anthropology that meaning is not only found in overt discussion, but also can be identified through the web of connotations that core concepts automatically activate. Co-retweet network analysis identifies communities of shared attention on Twitter, revealing how climate change discourse on the platform is socially organized and how these discourse communities evolve over time. Fine-tuned RoBERTa classification detects and categorizes specific types of climate denial at scale, sorting hundreds of thousands of tweets into a taxonomy of claims to measure what kinds of skepticism dominate and how their prevalence shifts over time.

No single method tells the full story. Topic models reveal what is being discussed; embeddings reveal how concepts are implicitly related; networks reveal who is connected and how communities form; and classification reveals which specific claims circulate and at what volume. Together, they construct a multi-layered portrait of an ecosystem of shared meanings that no single approach could produce alone.

Key Findings

1. Different media, different skepticisms. Conservative think tanks and Fox News frame climate change policy as an antagonist to free-market flourishing; they focus their effort on opposing regulations, carbon taxes, and international agreements. But climate skeptical blogs represent a discursive shift: their language is more conspiratorial, more emotionally charged, and engages in a view of general hostility towards scientific authority while promoting a wide-ranging skepticism.

2. The semantic world of climate skepticism is distinctive. Word embedding analysis reveals that in climate skeptical sources, "climate change" lives in a semantic neighborhood dominated by terms of deception, alarmism, and institutional corruption—a markedly different conceptual landscape than mainstream or liberal media, where climate change clusters with scientific and policy vocabulary.

A telling contrast: In CNN's language model, the nearest neighbors to "climate change" are greenhouse gas, environmental, and emissions. In the blog corpus, they are human caused, man-made, and agw (Anthropogenic Global Warming)—terms that frame the science itself as the object of contestation.

Semantic Worlds of “Climate Change”

Word2Vec models trained on each source reveal what concepts live nearest to “climate change” in that discourse. Select a source to see its semantic neighborhood.
Science / Physical
Contestation / Skepticism
Policy / Action

3. The "Climategate" pseudo-scandal crystallized a counterpublic. The November 2009 leak of climate scientists' emails catalyzed the formation of a durable, affectively bound online community. Network analysis of Twitter co-retweet patterns shows the climate skeptic network expanding from 17 core accounts before Climategate to 500 during the event, with community structure persisting long afterward.

4. The Twitter counterpublic is stable and polarized. Co-retweet networks from 2010 to 2020 reveal a remarkably consistent bifurcated structure: climate skeptics and climate advocates occupy separate, densely connected clusters with few bridges between them. Network transitivity remains below 0.5, confirming that audiences boosting skeptic content and those boosting scientific content are structurally partitioned.

5. Climate denial is ritualized and event-driven. Discussion in the skeptic counterpublic follows a pattern of stochastic bursts—rapid mobilization in response to climate events, IPCC reports, or political actions—followed by quieter periods. These bursts are powered by anger and disgust (confirmed by emotion classification), and function as collective rituals that reaffirm group identity and reinforce conspiratorial narratives.

Anatomy of Denial

A fine-tuned RoBERTa transformer model classified 260,000+ climate skeptical tweets into specific claim types. Block size reflects tweet volume. Hover for details.
It’s not real
It’s not us
Science / movement is corrupt
Policy / resilience
Attacks on the credibility of the climate movement overwhelm every other category combined—climate denial on Twitter is less about counter-science and more about conspiratorial rejection of authority and highly emotional partisan opposition

The Arc of Climate Skepticism

1970s–1990s
Anti-environmentalism emerges as a conservative movement defending anthropocentric, pro-carbon free-market neoliberalism. Environmental science is framed as covert propagation of an illegitimate "nature religion."
2000s
Conservative think tanks and Fox News build an institutional apparatus for skepticism, centering criticism of figures like Al Gore and recurrent narratives of scientific overreach. Climate skeptical blogs emerge as a grassroots counter-intelligentsia.
Nov 2009
Climategate. Leaked emails from the Climate Research Unit trigger a massive mobilization. The climate skeptic Twitter network explodes from a handful of accounts to hundreds, establishing durable community structures.
2010s
The skeptic counterpublic stabilizes into a persistent, polarized echo chamber on Twitter. Discussion shifts from coherent anti-environmentalist ideology toward diffuse, conspiratorial rejection of scientific authority.

Interactive: Network Evolution

The visualization below shows how the climate skeptic co-retweet network on Twitter evolved around the Climategate scandal. Each node is a Twitter account; connections indicate shared audiences (users who retweeted both). Colors represent detected communities. Click "Yearly Evolution" to see the network grow from 2009 to 2019.

Before Climategate (Oct 2009)
17
Accounts
13
Connections
5
Communities
0.096
Density
A sparse, disconnected network of 17 accounts sharing climate skeptic content. Small clusters orbit independently with no central coordination. This is the baseline before the leaked emails transformed the landscape.

Network Growth Over a Decade

From 313 accounts in 2009 to 8,344 in 2019 — the climate skeptic co-retweet network grew 27x in ten years, accelerating sharply after 2016.
The number of unique accounts in the co-retweet network grew steadily from 313 in 2009 to over 8,300 by 2019. Growth accelerated sharply after 2016 as climate skepticism merged with broader populist and conspiratorial movements.
Interactive network visualization. Use the play button and timeline dots to explore the Climategate phases. Hover over nodes for details.

What Makes This Research Distinctive

Most computational studies of climate skepticism analyze a single platform, apply a single method, or capture a narrow window of time. This study does something structurally different: it traces climate skepticism across four institutional contexts—think tanks, cable news, blogs, and social media—using four complementary computational methods on a unified twelve-year timeline.

This matters because climate skepticism is not a single phenomenon. The policy-focused opposition of a Heritage Foundation white paper, the conspiratorial engagement in the comments section of a Watts Up With That blog post, and the affective one-liners of a viral tweet are all “climate denial,” but they operate through different logics, in different institutional settings, for different audiences. By placing all four contexts on the same analytical timeline, this research can show how ideas originate in elite institutional settings, migrate through media intermediaries, and take on a life of their own as bedrock ideologies for durable grassroots communities online—a pipeline that is invisible from within any single platform.

Why It Matters

This research has three implications for understanding and responding to climate misinformation:

Beyond "information deficit." Climate skepticism cannot be resolved simply by providing better scientific information. The movement is sustained by affective bonds, identity performances, and ritualized collective action that operate independently of factual claims. Interventions must reckon with the emotional and communal dimensions of climate denial.

Platform structure matters. Social media platforms provide the infrastructure for rapid counterpublic formation and maintenance. The stability of the skeptic network over a decade—with the same core accounts maintaining centrality—suggests that platform-level interventions, not just content-level fact-checking, are necessary.

The conspiratorial drift. The evolution from structured anti-environmentalist ideology to diffuse online conspiracism represents a broader pattern in American public life. Climate skepticism's trajectory from policy disagreement to wholesale rejection of scientific authority offers a template for understanding similar dynamics in public health, election integrity, and beyond.

This brief draws on ongoing research by J. Connor Balch (Just Horizons Alliance), currently being prepared for peer-reviewed publication. The underlying study was developed with advisory input from Wesley J. Wildman (Computational Social Science), Neha Gondal (Social Network Analysis), and Christopher Wells (Emerging Media Studies) at Boston University. The climate denial claim classification builds on the CARDS framework (Computer-Assisted Recognition of Climate Change Denial and Skepticism) by Travis G. Coan, Constantine Boussalis, John Cook, and Mirjam O. Nanko.