Under the (neighbor)hood: Hyperlocal Surveillance on Nextdoor

Title of the Paper

Under the (Neighbor)hood: Hyperlocal Surveillance on Nextdoor

Bibliographic Information

• Subject Area: Social Computing, Privacy, and Community Surveillance
• Keywords: Privacy, Social Media, Online Communities, Urbanization, Methods, Theory, Qualitative Methods, Quantitative Methods

Research Background and Problem Statement

• Identified Issues or Challenges:

  • Nextdoor, as a neighborhood social media platform, faces discussions about the potential harm of community surveillance content, particularly in minority communities and neighborhoods undergoing gentrification.
  • Public disclosure of surveillance information about outsiders may foster social exclusion, negatively impact vulnerable groups, and promote vigilant-style community surveillance behaviors.

• Significance:

  • As a hyperlocal platform, Nextdoor addresses practical issues related to urbanization, crime, and community surveillance, influencing offline community behaviors and social tensions.

• Research Motivation and Related Work:

  • Current research focuses on the impact of user-generated content on Nextdoor on community dynamics and explores how it constructs and sustains digital communities.
  • This study aims to fill the gap in existing literature by providing an in-depth analysis of the quantitative and semantic relationships in community surveillance content.

Proposed Solution

• Proposed Approach:

  • Design and implement a privacy-preserving data collection process to extract data from public Nextdoor posts for identifying community surveillance patterns.
  • Develop a qualitative coding manual and utilize large language models (LLMs) for automated annotation and analysis of the data.

• Innovations:

  • Propose a scalable and empirically validated taxonomy of community surveillance patterns.
  • Develop a data annotation process that applies LLMs to qualitative analysis of social platform texts, offering new directions and technical frameworks for future research.

• Implementation Steps and Techniques:

  • Collect posts from Nextdoor communities in the Atlanta area (timeframe: March 2022 to March 2023).
  • Use filtering keywords and semantic models (S-BERT) to identify surveillance-related content.
  • Manually annotate the data and employ GPT-4 for large-scale automated data annotation.
  • Conduct Multiple Correspondence Analysis (MCA) to visualize patterns and trends in community gentrification.
  • Correlate gentrification indices with the distribution of surveillance-related posts for further exploration.

Research Findings

• Specific Outcomes:

  • Developed a taxonomy of community surveillance posts, including three main types: incident reports, community norm-establishing posts, and information dissemination posts.
  • Findings indicate a significant relationship between community surveillance content and gentrification trends:
    • Communities undergoing gentrification are more likely to post reports on specific crimes.
    • Communities in the "exclusive" stage of gentrification are more inclined to post about community norms or information.
    • Posts from non-gentrified communities are more emotionally charged, containing both positive and negative content.

• Advantages:

  • Provides empirical and qualitative support for understanding how hyperlocal social platforms influence community dynamics.
  • Extends and validates existing academic theories, such as social capital theory of community behavior and issues of digital belonging, using Nextdoor data.

• Experiments and Evaluation:

  • MCA analysis revealed two dimensions representing "specific crimes" and "community concerns," uncovering associations between different community types and content trends.
  • Validated the reliability of GPT-4 annotations, achieving an average Cohen's κ above 0.6.

• Limitations and Future Directions:

  • Limitations:
    • Simplification of the gentrification index may reduce the granularity of the analysis.
    • Data selection may introduce bias, such as prioritizing posts with higher interaction.
    • The MCA method did not include control variables, leading to results based primarily on correlation rather than causation.
  • Future Directions:
    • Expand the study to other cities and platforms to compare surveillance patterns across different social contexts.
    • Explore how to design fairer mechanisms for managing community dynamics, such as enhancing the representation of local community diversity.
    • Investigate the "hierarchy of needs" in community behaviors during gentrification and assess its universality.

Conclusion

This study reveals the complexity of community surveillance behaviors on the Nextdoor platform, exploring the relationships between data privacy, community norm formation, and gentrification trends. By establishing a taxonomy and qualitative analysis framework, this work provides new perspectives on how hyperlocal social platforms influence offline social relationships. It also offers profound insights into issues of social exclusion and digital belonging, with important implications for improving platform design and promoting community diversity.