Agency Aspirations: Understanding Users’ Preferences And Perceptions Of Their Role In Personalised News Curation

Title of the Paper

Agency Aspirations: Understanding Users’ Preferences and Perceptions of Their Role in Personalised News Curation

Paper Information

• Subject Area: Research on user behavior and perceptions in personalized news recommendation systems
• Keywords: Personalization, user experience design, interaction design, algorithmic news recommendation, user agency, filter bubble

Research Background and Problem Statement

• Issues and Challenges:

  1. The widespread use of algorithmic recommendations in personalized news delivery has transformed the way news is curated and consumed, leading to a lack of user awareness regarding their active role in news curation.
  2. Personalized recommendations may result in phenomena such as "filter bubbles" and "information fragmentation," exacerbating societal polarization.
  3. News curation requires balancing traditional journalistic values with personalized needs through the interaction of editors, algorithms, and users.

• Significance of the Research: In the digital media era, algorithmic recommendations have a profound impact on news dissemination. Exploring ways to empower users in the personalized news process to mitigate negative effects like filter bubbles is crucial for enhancing user experience and upholding the democratic values of journalism.

• Research Motivation and Related Work: Inspired by recent studies on user trust, transparency, and sense of control in algorithmic systems, this research aims to explore how users understand and exercise their agency in personalized news recommendations, addressing gaps in the study of user behavior and algorithmic interactions.

Solution

• Proposed Methods and Approach:

  1. Designed and tested an interactive news recommendation system prototype, named NAIRS, to explore users' perceptions of agency and control in the personalized news recommendation process.
  2. Conducted a mixed-methods study, including extensive surveys, in-depth interviews, and user interaction research with the designed prototype.

• Innovations:

  1. Employed a "provotype" as a provocative design tool to stimulate user reflection and behavior.
  2. Conducted a comprehensive analysis of transparency, user agency, and the interplay between algorithmic and editorial roles in personalized recommendations.
  3. Introduced the "News Personality Types" system to reveal algorithmic insights into user behavior, encouraging users to reconsider their news consumption habits.

• Implementation Steps and Key Techniques:

  1. Preliminary Surveys and Interviews: Conducted quantitative and qualitative research with 211 young users (aged 16-34) in the UK to understand their attitudes toward personalized news recommendations.
  2. Prototype Design: Developed the NAIRS system to capture user interaction data with news content and allocate personalized news based on their behavior.
  3. User Study (n=16): Engaged participants in two rounds of using NAIRS, exploring their reactions to transparency, agency, and personalized recommendations through an "override" mechanism for algorithmic/editorial news types.

Research Findings

• Key Outcomes:

  1. Users generally expressed a desire for greater agency, believing that control and choice contribute to a sense of security, trust, and autonomy.
  2. While many users appreciated having control, some were reluctant to actively intervene in practice, revealing a gap between behavior and intent.
  3. Assigning users to "News Personality Types" was seen by some as a self-discovery tool, with algorithmic analysis prompting users to redefine their self-perception.
  4. Users had fundamental demands for transparency, viewing an understanding of recommendation system mechanisms as a prerequisite for exercising agency.

• Comparison with Existing Solutions and Advantages: Compared to traditional news recommendation systems, NAIRS significantly enhanced user-system interaction and trust through its transparency (displaying the process of user profiling) and intervention design (allowing users to modify their profiles).

• Experiment and Evaluation Results:

  1. Transparency increased user acceptance of system adjustments, with most users satisfied with the system's reassignment of personality labels.
  2. The system repeatedly validated the adaptability between algorithmic recommendations and changes in user behavior, highlighting the potential of dynamic analysis and feedback mechanisms.
  3. However, the current design did not fully address user concerns about the workload of interventions or fears regarding the misuse of personal data.

• Limitations and Future Directions:

  1. The sample primarily consisted of young users with high educational backgrounds. Future research should include a broader demographic to enhance generalizability.
  2. The current design limited the diversity and iteration of user interactions. Future work should focus on developing more advanced personalization features.
  3. Ethical issues related to recommendation algorithms, such as editorial bias and user data privacy, require further exploration.