Impacts of Human and AI Moderation on Democratic Listening Online

The next speaker in this ECREA 2024 session is Shota Gelovani, who shifts our discussion further to the theme of democratic listening: the scrutiny and constructive discussion of statements by other citizens in a democracy. This can happen also between dissenting individuals, and may lead, if not to the removal of differences, then at least to partial consensus and an enlightened dissent.

Analytical listening (understanding others’ views) and critical listening (identifying errors in others’ judgments) are especially central to this, and democratic listening will also involve giving listening signals, substantive engagement with others’ claims, taking perspectives, and generating a feeling of being heard. To explore these, the project created several versions of an online forum which variously featured listening-oriented buttons (LOB), listening-oriented human moderation (LOHM), and listening-oriented AI moderation (LOAIM), and assessed listening quality in 40 scheduled 30-minute online discussions on climate change topics; it also ran pre- and post-surveys to test for cognitive and/or affective depolarisation effects.

AI as well as human moderation did produce somewhat greater levels of democratic listening, with AI very slightly ahead; group size also emerged as a significant factor in affecting the quality of group discussion, with groups larger than 15 quickly becoming less efficient.