Good deliberation regardless of mode? An experimental comparison of automated online moderation, human online moderation, and face-to-face moderation

Does the mode of deliberation – face-to-face with human moderators, online with human moderators and online with an automated moderation – influence opinion formation and learning in a deliberative mini-public? We study a Citizens’ parliament which was organized according to the Deliberative poll® model. Participants (n=671) were recruited through a stratified random sample of the Finnish population (n=30,000). They were randomly allocated into one of the three treatment conditions. In the Face-to-face condition, participants discussed face-to-face with human moderators. In Online human moderator, participants discussed in an online meeting (via Zoom) with human moderators. In Online automated moderator, participants discussed in an online platform with automated moderation. All discussions took place in small groups of around ten people. The topics were based on four citizens’ initiatives, two of which were about drug use policies and two about fuel pricing and taxation. Apart from the meeting mode, the procedures were held constant. Participants received briefing materials and rules of discussion prior deliberations, and during deliberations they could pose questions to experts. Each participant completed four surveys, of which three were administered before deliberations (t1-t3) and one after (t4). Surveys included items on opinions on the discussed topics, democracy preferences, political trust and efficacy, affective polarization and political knowledge. We hypothesize that: Participants’ opinions de-polarize and on average become more progressive (H1); Opinion changes are not different in the three modes of deliberation (H2); Participants’ issue knowledge increases in the course of deliberation (H3); Knowledge changes are not different in the three modes of deliberation (H4).