Social Media Feed Elicitation

TL;DR

Through feed elicitation interviews, an interactive method is introduced that guides users through identifying gaps and articulating their preferences to better author custom social media feeds, finding that participants significantly prefer the feeds produced from their elicited preferences to those produced by users manually describing their feeds.

Abstract

Social media users have repeatedly advocated for control over the currently opaque operations of feed algorithms. Large language models (LLMs) now offer the promise of custom-defined feeds--but users often fail to foresee the gaps and edge cases in how they define their custom feed. We introduce feed elicitation interviews, an interactive method that guides users through identifying these gaps and articulating their preferences to better author custom social media feeds. We deploy this approach in an online study to create custom BlueSky feeds and find that participants significantly prefer the feeds produced from their elicited preferences to those produced by users manually describing their feeds. Through feed elicitation interviews, we advance users' ability to control their social media experience, empowering them to describe and implement their desired feeds.

Figures (7)

• Figure 1: The elicitation interview proceeds in three main stages, meant to direct user intention towards the different main areas of interest for their specification
• Figure 2: The feed elicitation process produces thorough specifications with clarifying details about the feed intention, preferred quality attributes, ideal types of content, and content to remove or downrank. (Bolding added by researchers for emphasis of notable additions.)
• Figure 3: Each participant completes two different types of feed specifying activities, one being a baseline. Then they perform various quality rating and ranking tasks to evaluate overall quality and preference.
• Figure 4: Participants overall prefer the elicitation interview over the baseline and show no significant preference either way for the structured manual specification versus the baseline. However, participants who completed the interview lean towards more extreme feelings in both directions, feeling strong preferences for and against the interview condition at higher rates. Over a third of participants had no preference between the two manual specifications, whereas closer to a quarter had no preference between the interview and baseline feeds.
• Figure 5: Participants approved the posts at a modest but significantly higher rate for feeds created by the elicitation interview as compared to the manual specification.