Paper published in PLOS One Journal!
Adaptive political surveys and GPT-4
The paper "Adaptive political surveys and GPT-4: Tackling the cold start problem with simulated user interactions" has been published in the interdisciplinary journal PLoS One. The paper looks at how well GPT-4 can mimic user responses in a political survey and how such synthetic data helps to train statistical models in adaptive questionnaires. Doing so, it provides a method for solving the coldstart problem - one of the main issues in this field. The paper applies this method to an adaptive version of the Swiss Voting Advice Application Smartvote, and shows that it can significantly increase the recommendation accuracy for early users.
Congratulations to Fynn Bachmann, Daan van der Weijden, Lucien Heitz, Cristina Sarasua and Abraham Bernstein for the publication, which can be found here.