https://peterenns.org/ Professor of Government & Public Policy, Cornell University; Robert S. Harrison Director of https://socialsciences.cornell.edu/; Co-founder, https://www.verasight.io/
Peter K. Enns
Loading...
This Could Be the Worst Ebola Outbreak in History: I added two figures - from the same article - that extend the x-axis, which I think more clearly illustrate how much faster confirmed cases are accelerating that prior outbreaks. www.nytimes.com/interactive/...
As companies begin to sell synthetic (AI-generated) survey responses, new research with @gelliottmorris.com and Ben Leff shows why you should be extremely skeptical of this approach:
www.verasight.io/reports/can-...
Peter K. Enns
Verasight ran four experiments on AI "digital twins." Synthetic samples can match polls on Trump approval and the generic ballot, but fail systematically on almost everything else.
The new ANES Board of advisors met this week in Ann Arbor to begin planning the 2028 study. It is a fantastic group and the PIs are very grateful for their dedication to the project and to the discipline. Once more unto the breach, dear friends!
Four studies. One year. The same conclusion.
We tested whether AI can replace survey respondents, from simple persona prompts to a from scratch steelman pipeline. Synthetic samples are predictions, not measurements.
New review from @gelliottmorris.com, @peterenns.bsky.social, and Benjamin Leff.
Never-before-seen data from top Democratic pollster Stan Greenberg! Download reports and banner books from his US work for Bill Clinton and Al Gore, as well as his political surveys in Bolivia, Israel, and Vienna!
https://bit.ly/4dUp3Rd
We built an AI "digital twin" and tested it against human survey respondents.
Across 16 questions it still missed by 6.6 points on average. The gap held across every model and method we tried. New study from @gelliottmorris.com, @peterenns.bsky.social, and Benjamin Leff. Link in reply.
What if polling organizations aren’t just passive recorders of political sentiment but also shape the agendas of other policy institutions?
Qian Zhang @peterenns.bsky.social and I ask this question in @psjeditor.bsky.social (open access)
onlinelibrary.wiley.com/doi/10.1111/...
Hey look yet ANOTHER study, this one from @gelliottmorris.com, @peterenns.bsky.social, and Ben Leff showing that digital twins don't work www.verasight.io/reports/can-...
4/6 🧵 In Sociological Science Enns questions recent findings that policy responds more to the preferences of more affluent citizens. Restricting attention to cases of disagreements under certain conditions can generate a spurious estimate of unresponsiveness to the poor doi.org/hb2f6c
onlinelibrary.wiley.com
Polling organizations, like other policy actors, must prioritize certain issues. We argue that, for normative and financial reasons, pollsters prioritize issues that are viewed as important by other ...
Verasight ran four experiments on AI "digital twins." Synthetic samples can match polls on Trump approval and the generic ballot, but fail systematically on almost everything else.
www.verasight.io
Verasight
Verasight
Nick Valentino
The situation in eastern Congo and Uganda combines some of the most dangerous aspects of the 2014 and 2018 Ebola outbreaks.
