Happy to report that this article is now available: doi.org/10.1017/S000.... See Tyler's thread below for more detail on the associated R package and interactive online app.

"Place Types." Excited that this paper with @sborwein.bsky.social is now available. We develop a new typology of place types in Canada at a fairly fine-grained level of geography, and then show how these place types relate to Canadian politics. doi.org/10.1017/S000...

Very excited to share our paper on the vote intention dataset in @cjps-rcsp.bsky.social.

See the earlier thread for more on the Shiny app, R package, and descriptive analyses of demographic divides in vote intention. We'll keep updating this resource—stay tuned!

📄 www.cambridge.org/core/journal...

Special thanks to @anne-imz.bsky.social, the grad students/faculty involved with the CSDC at McGill University, the @clessn.bsky.social research team at Université Laval, and our discussants at past conferences.

We are very grateful for the constructive feedback and thought-provoking questions! 🙂

As AI’s capabilities expand, opportunities to study the dynamic nature of citizens' risk perceptions and their effects on human behaviour appear boundless.

We hope our dread-controllability framework motivates deeper theoretical and empirical exploration of how citizens perceive AI and its risks.

We validate these measures by regressing each scale on covariates linked to AI attitudes.

As shown below, key cross-national predictors of AI dread and controllability concerns include trust in scientists, conspiracy thinking, and beliefs that technological change will harm one’s job prospects.

The items for the AI Dread and AI Controllability Concern Measures.

Using data from 🇨🇦 and 🇯🇵, we develop four-item dread and controllability measures and identify the single best item to use for researchers with limited survey space.

The scales are highly reliable and largely uncorrelated (r = 0.05 in Canada, 0.01 in Japan). We summarize these new measures below.

Chess robot grabs and breaks finger of seven-year-old opponent Moscow incident occurred because child ‘violated’ safety rules by taking turn too quickly, says official

For those interested in some real-world developments surrounding these ideas, two Guardian articles linked below are quite fascinating:

www.theguardian.com/sport/2022/j...

www.theguardian.com/technology/2...

We suggest that controllability concerns stem from:

(a) unease about AI appearing outright out-of-control akin to Frankenstein's monster type scenarios;

(b) uncertainty over who controls it, seen in debates about AI capitalism and authoritarian regimes using AI to push revisionist narratives.

Image source: https://theweek.com/political-satire/1022837/ai-takeover

If you're like me, some ideas are better contextualized through memes or playful references, like this creative cartoon that loosely illustrates a key component of our dread argument.

Image credit: Joe Heller.

We theorize that citizens’ dread of AI stems from:

a) Its impact on employability, inequality, and the disappearance of professions.

b) Its capacity to match or surpass human intellect/ undermine human initiative and independent thinking.

c) Its potential to pose existential threats to humanity.

Dread captures the perceived magnitude of risk associated with AI, while controllability reflects the perceived ability to manage its development and consequences.

We propose a new framework centered on these dimensions and introduce two measures: the AI Dread and AI Controllability Concern Scale.

Studies often link AI attitudes to demographics, predispositions, or framing. Yet, the factors structuring these beliefs remain contested.

We argue that a growing literature implicitly highlights how citizens’ views of AI and its risks stem from a sense of dread and a perceived lack of control.

Image and news article source: https://www.aljazeera.com/economy/2024/2/12/how-ai-is-used-to-resurrect-dead-indian-politicians-as-elections-loom

Image and news article source: https://www.theguardian.com/us-news/2024/feb/26/steve-kramer-admits-he-commissioned-robocall-ai-biden-new-hampshire

Image and news article source: https://www.nytimes.com/2022/09/02/technology/ai-artificial-intelligence-artists.html

AI has immense potential to drive innovation, but recent news highlights its complex societal role—from attempts to “resurrect” politicians to sway elections, to using AI to suppress voter turnout or win art competitions. These issues raise a key question: how do lay citizens view AI and its risks?

How do ordinary citizens think about AI and its associated risks cross-nationally—and how can we measure it?

I’m thrilled to share a new paper in Journal of Risk Research with @tylergirard.bsky.social, Mathieu Turgeon, Yannick Dufresne, Takeshi Iida & Tetsuya Matsubayashi.

doi.org/10.1080/1366...

This new R package makes it easy to extract weighted annual estimates of vote intention for the major parties going back to the 1940s. For instance, here's annual vote intention for the Liberals, with a red line marking where they currently stand in the polls according to the 338 aggregator.

If you have any questions or would like to discuss this further, please feel free to reach out to me or Jack.

Let's dive into the data! #CdnPoli #Election2025

After many years of weekly Zoom hackathon sessions, we're excited to release this!

We'll continue updating the resource (soon with issue attitudes) and hope it helps illuminate key demographic trends, regional shifts, and electoral changes from postwar realignments to contemporary divides.

Special thanks to Dave Armstrong for leading the way with the R package and Shiny App! The Shiny is available here: quantoid.shinyapps.io/cvpa_app/

We also hope the data will be a helpful tool for teaching and reaching the wider public. It allows users to select a specific demographic variable, choose parties, and view trends in vote intent/choice or explore the cross-classification of two variables, such as gender and education (among others)

The R package is available for download here: github.com/davidaarmstr....

The first function, wtd_vote, creates weighted estimates of party support for specified demographic groups, while gap_analysis generates statistical estimates of the difference in support among demographic group members for each party. Each figure above was created using these functions.

In addition, we’ve created a new R package, Canadian Voting and Policy Attitudes ("CVPA"). The CVPA package focuses on providing weighted party support by demographic groups over time. It contains the raw data and two functions designed to be particularly helpful for researchers.

Special thanks to Alex Cooper at Queen's University for making the dataset publicly available on the Canadian Opinion Research Archive! Those interested can download the data here: doi.org/10.5683/SP3/...

The dataset also includes age, education, gender, occupation, religion, union membership, language (not pictured), province, region, and community size variables. It also provides weights, survey mode, and interview dates, offering new insights into responsiveness to events and survey methodology.

And the community size divide...

We believe the dataset is well-equipped for both descriptive and rigorous analyses of changes in Canadians' vote intention over time. Below, we highlight the long-term development of three key demographic divides in vote intention: gender and education.

Thrilled to introduce the Canadian Vote Intention Dataset—a harmonized database of vote intention, choice, demographics, and place variables from 1945–2022. These data include 1M+ responses from 680 surveys by Gallup, Environics, Pollara, and CES/CDEM. Details below about the variables included.

With the federal election approaching, I’m excited to share insights on how Canadians' vote intentions have shifted since 1945. This forthcoming CJPS paper, with @jacklucas.bsky.social, Dave Armstrong, and @eplusgg.bsky.social, features a public dataset, R package, and Shiny app—more details below!