News Nutritionist: Leveraging LLMs for improved news curation in social media feeds

Abstract

Younger news consumers are increasingly turning to social media platforms to get the information they need. However, the news they consume there can be misaligned with what they want to see because social media algorithms award engagement rather than intent. We introduce News Nutritionist, a system that enables people to create intentional news feeds on the social media platform Bluesky through conversation with an AI news curator. Through the combination of a conversational on-boarding and feed adjustment chatbot that lives in users direct messages, users can create custom news feeds that match their needs and expressed preferences more than traditional social media algorithms. We evaluated the system with 10 young adults in a two phase, one week study. We find that users created news feeds that included more content they felt they should see but did not find as engaging as other content available on social media. We also find that although users felt negatively about news content, they still wanted the system to surface it, suggesting that intentional curation can capture desired content that engagement-based algorithms miss. Additionally, we found that there was a tradeoff in our conversational interface design between reducing the cognitive burden of creating an intentional feed and leading users decision making. Overall, our work highlights how intentional social media feed curation can improve young news consumers experience with news content on social media, while highlighting how conversational design choices can shape, and sometimes restrain, the feeds users build.