Keywords: deliberative polling, generative ai, meta

A deliberative mini-public relies on a randomly chosen, representative sample of participants, and therefore understanding sample selection and recruitment strategies are key. Citizens who agree to participate might be systematically different than those who get invited and choose not to participate, and recruitment typically involves multiple invitation waves and extra efforts to access hard to reach populations. It is an open area of research as to the types of individuals who accept invitations to participate in deliberative mini publics, and how recruitment strategies influence the types of participants. Do the descriptive characteristics of participants vary across recruitment waves? Do we see differences in participation patterns across populations that were easier or more difficult to recruit? We use novel data from the 2023 Meta Community Forum on AI Chatbots, which brought together 1541 participants from Brazil, Germany, Spain, and the United States to participate in a deliberative polling exercise. We first provide descriptive data of the demographics of participants across recruitment waves, to demonstrate which individuals are more or less easily recruited. We then look at how recruitment waves affect participation, measured by speaking, interrupting, and voting behavior in the session.