Use a large language model to offer users rephrased comments to encourage constructive conversations.

To improve the civility of discussions.

Argyle et al. (2023) conducted an online experiment where pairs of participants discussed controversial topics. Some of the participants had access to an AI assistant that suggested rephrasing comments to either 1) restate the message of their conversational dialogue, implying understanding; 2) validate the other participant’s perspective; or 3) make the message more polite. Access to such an AI-assistant significantly improved their conversation partner’s reported conversation quality, and decreased a divisiveness metric created by the authors. Effect sizes estimated conservatively are in the 2.5-5% range. It is important to note that it is unclear whether these results would generalize to scenarios where knowledge about these AI-tools (and that people are using them online) is widespread -- especially since participants in the “receiving” end (i.e., whose conversational partner had access to the chat bot), were not aware of it; and since the study never considered the case in which both participants had access to the AI assistant.

Improving the quality of online conversation on divisive topics can help decrease polarization and increase the well-being of users.