Probing a Community-Based Conversational Storytelling Agent to Document Digital Stories of Housing Insecurity

Document Title

Probing a Community-Based Conversational Storytelling Agent to Document Digital Stories of Housing Insecurity

Document Information

• Subject Area: Artificial Intelligence, Community Design, Story Documentation in Social Movements, and Housing Insecurity.
• Keywords: Community Design, Conversational Interface, Housing Insecurity, Narrative Systems, Social Movements, Digital Stories, Artificial Intelligence, Cultural Adaptability, Multilingual Support.

Research Background and Problem Statement

• Problems and Challenges:

  • Digital stories are increasingly important in building social movements, yet their documentation faces challenges such as resource scarcity, volunteer attrition, and poor technological compatibility.
  • In the context of housing insecurity, victims often encounter threats, psychological stress, or blame narratives, which hinder the dissemination of authentic stories.
  • Collecting subset data is difficult to scale and insufficient to mitigate potential biases.

• Significance of the Research:

  • Effective story documentation can enhance public awareness within communities and contribute to resolving housing justice issues.
  • Overcoming the limitations of existing tools for narrative capture can uncover new social histories and provide data to support policy reforms.

• Motivation and Related Work:

  • Based on collaboration with the Anti-Eviction Mapping Project, the authors have utilized technology to expand multimedia story documentation during the pandemic.
  • Existing tools, such as user-generated content platforms and in-person interviews, are effective but limited by resources, volunteer availability, and adaptability.
  • Domestic and international literature and projects (e.g., Eviction Lab) highlight that expanding technological methods for storytelling can empower social movements but may also raise ethical risks and issues of digital inequality.

Proposed Solution

• Proposed Solution:

  • Design and test a community-supported conversational storytelling agent (CSA) to document digital stories of housing insecurity.
  • Integrate multilingual and multimedia support, leveraging natural language processing and machine learning technologies to enable users to document their stories anytime, anywhere.

• Innovative Aspects of the Solution:

  • Combine graphical design with traditional conversational interfaces to explore the applicability of "non-traditional conversational tools" while simulating "human-machine dialogue" for data collection.
  • Provide a "therapeutic narrative experience" through the CSA, functioning not only as a data documentation tool but also aiming to enhance users' psychological comfort.

• Implementation Steps:

  1. Create a visual design model for the CSA, including desktop and mobile versions.
  2. Design a CSA conversational prototype using the "Wizard of Oz" technique to simulate interviews.
  3. Conduct semi-structured interviews, inviting volunteers to use the CSA prototype, share their stories, and provide feedback on their experience.
  4. Analyze user feedback to assess the effectiveness and limitations of the CSA.

Research Outcomes

• Specific Outcomes:

  • Advantages:
    • The CSA provides a low-cost, convenient channel for documenting stories of housing insecurity, reducing the burden on volunteers.
    • Multilingual support expands story capture for individuals with limited English proficiency.
    • The "non-face-to-face" feature eliminates shame or social pressure during the storytelling process.
    • 24/7 availability makes storytelling more flexible.
  • Experimental and Evaluation Results:
    • Collected feedback from 25 users from diverse backgrounds in U.S. cities, including 5 storytellers, 4 interview recorders, and 8 individuals with dual roles.
    • User feedback validated the usability of the CSA, with some users expressing a preference for machine guidance over interaction with strangers.
    • However, the CSA cannot replace real human interviews, particularly in terms of emotional connection and deep resonance.

• Limitations and Future Directions:

  • Issues of machine bias, such as implicit gender or language biases, may arise during diverse interviews.
  • Lack of emotional support capabilities; automated interviews may cause emotional detachment or discomfort when addressing users' painful experiences.
  • Future research should explore hybrid models combining CSA with human interviews to enhance the community's role with technological support.
  • Design improvements should focus on building trust, transparency, and compliance in the CSA to prevent improper use of data.

Conclusion and Implications

Feedback from community volunteers and storytellers in this study highlights the potential contributions of the CSA to the field of digital storytelling while emphasizing the limitations of digitization and automation. Future tool design should prioritize "human touch" and the needs of sustainable community development, integrating technological solutions with local socio-cultural practices.