Stances on climate-action narratives
Stances on climate-action narratives
Embedded Files
Methods
Methods
• 296 million tweets by 99,274 scholars were scanned.
• GPT-4 produced year-specific climate-policy dictionaries; GPT-3.5-turbo then classified up to three English tweets per author / month as pro, anti, neutral (stance scale −1 … +1) and as one of two supportive narratives: Techno-optimism vs Behavioural-adjustment.
• The global mean stance for Climate Action is +0.085 (95 % CI [0.083, 0.086]).
Key patterns
Key patterns
Gender: Male scholars are markedly more techno-optimistic, whereas women show a (small) tilt toward behavioural solutions.
Field: STEM is both the most climate-supportive and the most techno-optimistic; social sciences and humanities lean more on behavioural change.
Reach × Expertise: High-reach non-experts display the weakest support for climate action; low-reach experts the strongest.
Institution rank: Scholars at top-100 universities are less pro-climate than peers at ranks 101–500.
Geography: U.S.-based academics score 0.071 on climate action—lower than the 0.090 average outside the U.S.
Cultural-liberal and economic-collectivist expression
Cultural-liberal and economic-collectivist expression
Methods
Methods
• Same corpus and sampling as Fig. 2.
• Two composite stances: Cultural Liberalism (race, immigration, abortion) and Economic Collectivism (welfare, redistribution, tax on the wealthy); scale −1 … +1.
• Overall means: Cultural = +0.043; Economic = +0.018.
Findings
Findings
Gender: Women post slightly more liberal (+0.050) and collectivist (+0.020) content than men.
Field: Social scientists are the most collectivist; humanities edge out others on liberalism.
Reach × Expertise: High-reach non-experts moderate both stances; low-reach experts are the most progressive.
Institution rank: Top-100 universities are more culturally liberal than all other tiers; economic collectivism shows no rank gradient.
Geography: U.S. scholars are more culturally liberal than non-U.S., but similar on economic collectivism.
Behavioural style online
Behavioural style online
Metrics (per tweet, 2016 – 2022)
Metrics (per tweet, 2016 – 2022)
Egocentrism share of first-person-singular words.
Toxicity Perspective-API probability (0–1).
Emotionality / Reasoning ratio of affective ÷ cognitive terms (LIWC).
Headline results
Headline results
Egocentrism: Highest in humanities (0.366) and among U.S. scholars; also rises with university rank and with high reach / low credibility.
Toxicity: Mean 0.044. Humanities, social sciences, and top-100 universities exceed the academic average; STEM and low-reach experts are lowest.
Emotionality: Females, STEM fields, and high-reach non-experts use the most affect-laden language; differences across rank are negligible.
Page updated
Google Sites
Report abuse