We develop an AI chat assistant that makes real-time, evidence-based suggestions for messages in divisive online political conversations. In a randomized controlled trial, we show that when one participant in a conversation had access to this assistant, it increased their partner’s reported quality of conversation and both participants’ willingness to grant political opponents space to express and advocate their views in the public sphere. Participants had the ability to accept, modify, or ignore the AI chat assistant’s recommendations. Notably, participants’ policy positions were unchanged by the intervention. Though many are rightly concerned about the role of AI sowing social division, our findings suggest it can do the opposite—improve political conversations without manipulating participants’ views.
Political discourse is the soul of democracy, but misunderstanding and conflict can fester in divisive conversations. The widespread shift to online discourse exacerbates many of these problems and corrodes the capacity of diverse societies to cooperate in solving social problems. Scholars and civil society groups promote interventions that make conversations less divisive or more productive, but scaling these efforts to online discourse is challenging. We conduct a large-scale experiment that demonstrates how online conversations about divisive topics can be improved with AI tools. Specifically, we employ a large language model to make real-time, evidence-based recommendations intended to improve participants’ perception of feeling understood. These interventions improve reported conversation quality, promote democratic reciprocity, and improve the tone, without systematically changing the content of the conversation or moving people’s policy attitudes.