Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Negotiating the Shared Agency between Humans & AI in the Recommender System

Mengke Wu, Weizi Liu, Yanyun Wang, Mike Yao

TL;DR

This paper addresses the problem of user agency in recommender systems amid information and power asymmetries. It introduces a dual-control mechanism that lets users govern both data collection and the degree of algorithmic tailoring via an Algorithm Outcome Control (AOC) slider, complemented by data-control options. In a between-subject experiment with 161 participants, the authors show that transparency alone can fail to enhance, and may even undermine, perceived agency, whereas coupling transparency with user controls—especially direct outcome control—significantly boosts perceived control and engagement. The work provides a proof-of-concept and practical design guidelines for more user-centered AI-driven content delivery, with implications for fairness, autonomy, and trust in RS.

Abstract

Smart recommendation algorithms have revolutionized content delivery and improved efficiency across various domains. However, concerns about user agency arise from the algorithms' inherent opacity (information asymmetry) and one-way output (power asymmetry). This study introduces a dual-control mechanism aimed at enhancing user agency, empowering users to manage both data collection and, novelly, the degree of algorithmically tailored content they receive. In a between-subject experiment with 161 participants, we evaluated the impact of varying levels of transparency and control on user experience. Results show that transparency alone is insufficient to foster a sense of agency, and may even exacerbate disempowerment compared to displaying outcomes directly. Conversely, combining transparency with user controls-particularly those allowing direct influence on outcomes-significantly enhances user agency. This research provides a proof-of-concept for a novel approach and lays the groundwork for designing more user-centered recommender systems that emphasize user autonomy and fairness in AI-driven content delivery.

Negotiating the Shared Agency between Humans & AI in the Recommender System

TL;DR

This paper addresses the problem of user agency in recommender systems amid information and power asymmetries. It introduces a dual-control mechanism that lets users govern both data collection and the degree of algorithmic tailoring via an Algorithm Outcome Control (AOC) slider, complemented by data-control options. In a between-subject experiment with 161 participants, the authors show that transparency alone can fail to enhance, and may even undermine, perceived agency, whereas coupling transparency with user controls—especially direct outcome control—significantly boosts perceived control and engagement. The work provides a proof-of-concept and practical design guidelines for more user-centered AI-driven content delivery, with implications for fairness, autonomy, and trust in RS.

Abstract

Smart recommendation algorithms have revolutionized content delivery and improved efficiency across various domains. However, concerns about user agency arise from the algorithms' inherent opacity (information asymmetry) and one-way output (power asymmetry). This study introduces a dual-control mechanism aimed at enhancing user agency, empowering users to manage both data collection and, novelly, the degree of algorithmically tailored content they receive. In a between-subject experiment with 161 participants, we evaluated the impact of varying levels of transparency and control on user experience. Results show that transparency alone is insufficient to foster a sense of agency, and may even exacerbate disempowerment compared to displaying outcomes directly. Conversely, combining transparency with user controls-particularly those allowing direct influence on outcomes-significantly enhances user agency. This research provides a proof-of-concept for a novel approach and lays the groundwork for designing more user-centered recommender systems that emphasize user autonomy and fairness in AI-driven content delivery.
Paper Structure (22 sections, 3 figures, 2 tables)

This paper contains 22 sections, 3 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Information Asymmetry
  4. Power Asymmetry
  5. Existing User Control in RS
  6. Opportunities for Shaping Power Dynamics
  7. METHODS
  8. Pretest
  9. Main Study
  10. Study Sample & Procedure
  11. Algorithm Outcome Control
  12. Flow Design
  13. Measurements
  14. Data Analysis
  15. PRELIMINARY RESULTS
  16. ...and 7 more sections

Figures (3)

  • Figure 1: Examples of Different Types of Agency. Perceptual Agency (Left): (a) Source Transparency, (b) Process Transparency; Behavioral Agency (Right): (c) Executing Permissions, (d) Specifying Preferences, (e) Providing Feedback.
  • Figure 2: Wireframe of the Recommendation Flows
  • Figure 3: Representative Finalized Webpages: (a) Landing Page, (b) User Data Control (UDC), (c) Algorithm Outcome Control (AOC), (d) Recommendation Result Page.