This figure illustrates academics' stances on three pivotal climate change policy topics across different academic groups and personal characteristics.

Method: We use the leading generative AI model GPT-4 to detect the topic of climate change policy within a sample of 138 million tweets made by 100,000 academics between 2016 and 2022. The average stance metric ranges from -1 (most expressive against climate action) to 1 (most expressive in favour of climate action), with the global average stance for Climate Action being slightly positive at 0.086. This metric captures the extent of support or opposition to climate action within the academic community, segmented by Gender, Fields of study, Twitter Reach combined with Academic Expertise, University Rankings, and Country.

Findings: Male academics tend to exhibit greater optimism towards technology solutions to address climate change compared to their advocacy for behavioural adjustment solution, which is also more prevalent among STEM than in social sciences or humanities. Academics from top-ranked U.S. universities, and those with higher social media reach and non-expertise in climate, are notably less supportive of proactive climate measures. For context, on average, academics are 10.7 times more expressive in their support for climate action compared to the general U.S. population.

This figure displays the distribution of academic stances on two significant socio-economic dimensions—Cultural Liberalism and Economic Collectivism—across diverse academic profiles. 

Methods: These metrics, derived using AI-driven text classification techniques similar to those used for climate policy topics, assess the average net stances from -1 (most expressive against) to 1 (most expressive in favour of). Cultural liberalism covers issues like racial equality, abortion rights, and immigration, while economic collectivism encompasses welfare state policies, income redistribution, and taxation. The analysis is  based on a dataset of 138 million tweets made by a balanced panel of 100,000 academics between 2016 and 2022.

Findings: average support for cultural liberalism among academics is slightly positive at 0.044, with a similar trend for economic collectivism at 0.019. This indicates a general progressive orientation in their online expressions. The figure stratifies data by Gender, Fields of study, Twitter Reach interacted with Academic Expertise, University Rankings, and Country, uncovering patterns such as greater economic progressiveness among female academics and among social scientists in general, compared to their counterparts in humanities and STEM. Additionally, academics from top-ranked U.S. universities and those with significant Twitter reach but non-expertise on the issue tend to express more conservative views on those issues. On average, academics are 6.2 times more expressive in their support for cultural liberalism and 23.3 times more expressive for economic collectivism compared to the general U.S. population 

This figure presents the average levels of three behavioural metrics—Egocentrism, Toxicity, and Emotionality/Reasoning.

Methods: quantified from tweets by a balanced panel of academics from 2016 to 2022, the figure uses data linking 100,000 Twitter profiles to their academic profiles. The figure is structured into three panels, each representing one of the behavioural characteristics:

1.  Egocentrism measures the average proportion of self-referential terms in expressions such as "I", "Me", "My", and "Myself".

2. Toxicity, calculated using Google’s Perspective API, rates tweets on a scale from 0 to 1, where higher values indicate greater toxicity. 

3. Emotionality/Reasoning assesses the ratio of affective to cognitive words, providing insight into the emotional versus rational content in tweets. 

Findings: Higher egocentrism levels are observed among academics from STEM fields. Academics affiliated with top 100 universities exhibit higher toxicity levels compared to their peers. STEM academics exhibit higher emotionality compared to those in humanities fields, who tend to use more cognitive words. On average, academics exhibit 78% of the egocentrism, 50% of the toxicity, and 88% of the emotionality compared to the general U.S. population.