When the Social Becomes Non-Human: A Study of Young People’s Perception of Social Support in Chatbots

Document Title

When Social Support Becomes Non-Human: Young People's Perceptions of Social Support from Chatbots

Document Information

• Subject Area: Human-Computer Interaction (HCI), Artificial Intelligence, Mental Health
• Keywords: Chatbots, Young People, Artificial Intelligence, Social Support, Privacy Trust, Mental Health, Informational Support, Self-reflection, Appraisal Support, Design Practices

Research Background and Issues

• Issues and Challenges:

  • Social support is crucial for young people's health and psychological well-being, yet many hesitate to seek support.
  • Existing research primarily focuses on human-to-human relationships, lacking an in-depth understanding of non-human social support, especially young people's perceptions and experiences in everyday contexts.
  • The proliferation of chatbots presents opportunities for non-human social support but also raises concerns such as privacy issues, ineffective responses, and potential psychological isolation.

• Significance and Research Motivation:

  • Due to the impact of the COVID-19 pandemic, young people's demand for online social support has significantly increased, and chatbots, as emerging tools for social support, hold great potential.
  • Investigating whether chatbots can play a positive role in mental health, informational support, and everyday social interactions.

• Related Work:

  • Previous studies have focused more on chatbots for elderly individuals and groups with specific mental health needs, with limited research on young people's everyday support experiences.
  • Current research urgently needs to further understand human-machine relationships, particularly the role and impact of privacy and trust in social support.

Solution

• Proposed Approach:

  • The authors use qualitative research methods to deeply investigate how young people perceive chatbots' performance in different types of social support.
  • This includes chatbots' performance in four traditional types of social support: appraisal support, emotional support, informational support, and instrumental support, while also exploring privacy and trust issues.

• Innovations:

  • Combining in-depth interviews with surveys to explore the complex experiences of chatbots providing social support in daily life.
  • Introducing Woebot (a mental health chatbot) and Ungbot (a prototype informational support chatbot) into the study to compare the support effects of social robots with different functionalities.

• Implementation Steps and Techniques:

  • Selecting participants aged 16 to 21, who first interact with Woebot for two weeks.
  • Conducting face-to-face in-depth interviews to explore participants' specific experiences.
  • Providing the Ungbot prototype for exploratory experiences in a laboratory setting.
  • Following up with participants via email two months later to assess long-term usage.
  • Analyzing interview data using NVivo and coding appraisal support, emotional support, etc., within a theoretical framework.

Research Findings

• Specific Findings:

  • Chatbots perform significantly in encouraging self-reflection and emotional expression.
  • Most young people find chatbots more accessible than human interactions, offering advantages such as anonymity and privacy protection.
  • Woebot helps users conduct self-assessments and guides them in identifying negative thought patterns and emotional states.
  • Information provided by chatbots is perceived as credible and efficient, suitable for solving everyday problems.

• Advantages:

  • Most participants believe chatbots lower the barriers to seeking social support, making them suitable for individuals uncomfortable with traditional social relationships.
  • Chatbots support self-reflection and provide informational support that experienced individuals may not obtain from friends or professionals.

• Experimental or Evaluation Results:

  • About half of the respondents reported that chatbots could provide emotional support, although the perception of effectiveness was sometimes limited by the non-human nature of the machines.
  • Long-term users of Woebot, mostly under life stress or psychological distress, acknowledged the value brought by the chatbot.
  • No major privacy issues were reported, and most participants expressed high trust in the chatbot and its data storage.

• Limitations and Future Directions:

  • The sample size is relatively small and primarily focused on digitally adept young people, with regional and cultural differences yet to be fully explored.
  • Long-term observation and larger sample sizes are needed to understand the development of user-chatbot relationships.
  • Recommendations include exploring how to enhance social connection capabilities and privacy control options in chatbot design.
  • Investigating chatbot privacy settings and user data flows to technically evaluate their impact on user trust.

Design and Practice Implications

• Chatbot design should consider diverse user needs, such as tailoring informational and emotional support functionalities to different scenarios.
• Social chatbots need stronger relationship-building and communication capabilities.
• Privacy and trust issues require special attention, with the creation of secure data protection environments being critical.
• Research is needed on long-term chatbot usage behavior patterns and habit formation processes to improve user experience.