Co-Authoring the Self: A Human-AI Interface for Interest Reflection in Recommenders

2025-10-09

AI Summary

In recommender systems, natural-language user profiles suffer from persistent misalignment between users’ self-perceived interests and AI-inferred interests, while offering limited support for user-initiated correction. Method: This paper proposes a dynamic, editable personalized interest summary interface that enables users to inspect, modify, and reflect upon AI-generated viewing preference profiles—facilitating human-AI collaborative interest articulation. The approach integrates natural-language user profiling, editable interface design, and seamless integration with an online recommendation system. We conducted an eight-week in-the-wild deployment involving 1,775 users. Results: The interface significantly increased user engagement and reflective interest revision behaviors, enhancing recommendation transparency, trustworthiness, and controllability. It provides empirical grounding and design insights for a novel human-AI collaboration paradigm wherein imperfect AI actively invites and supports user agency.

Technology Category

• Humans and AI: Intelligent User Interfaces
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Learning Preferences or Rankings

Application Category

• User Modeling, Personalization and Recommendation: Studies of user behavior, including longitudinal effects of personalized systems
• Economics, Online Markets and Human Computation: Economics and fairness of platforms and recommendation systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Natural language-based user profiles in recommender systems have been explored for their interpretability and potential to help users scrutinize and refine their interests, thereby improving recommendation quality. Building on this foundation, we introduce a human-AI collaborative profile for a movie recommender system that presents editable personalized interest summaries of a user's movie history. Unlike static profiles, this design invites users to directly inspect, modify, and reflect on the system's inferences. In an eight-week online field deployment with 1775 active movie recommender users, we find persistent gaps between user-perceived and system-inferred interests, show how the profile encourages engagement and reflection, and identify design directions for leveraging imperfect AI-powered user profiles to stimulate more user intervention and build more transparent and trustworthy recommender experiences.

Problem

• Addresses gaps between user-perceived and system-inferred interests
• Introduces editable AI-generated summaries for user reflection
• Leverages imperfect profiles to increase transparency and trust

Innovation

• Editable personalized interest summaries for users
• Human-AI collaborative profile for movie recommendations
• Leveraging imperfect AI profiles to encourage user intervention