Developing a Social Support Framework: Understanding the Reciprocity in Human-Chatbot Relationship

Research Background and Issues

• What problems or challenges did the authors identify?

  • Although numerous studies have demonstrated that chatbots can effectively provide emotional support and reduce user stress, existing research has significant gaps in the following areas:
    1. The types of social support have not been systematically categorized; most studies focus solely on emotional or informational support, neglecting other forms of support.
    2. Experimental designs lack long-term perspectives, making it difficult to establish genuine relationships between users and chatbots or to observe the effects of prolonged interactions.
    3. The bidirectionality and reciprocity of social support have not been systematically analyzed; most studies focus only on users receiving support from chatbots, rather than on users providing support or the dynamic processes of mutual interaction.

• Why is this issue important?

  • Chatbots are rapidly gaining traction as tools for psychological support, but it is crucial to clarify their roles in real-world contexts and their interaction mechanisms with users, particularly regarding the types and directions of social support. Enhancing understanding in this area can help optimize chatbot design and provide insights into ethical considerations.

• Research Motivation and Related Work

  • Current research indicates that chatbots can not only provide emotional and informational support but also serve as "artificial online friends," aiding in stress management, emotional communication, and mental health treatment (e.g., Replika). However, there is a lack of classification and analysis of the types and directions of social support exchanged between chatbots and users. This study aims to address this research gap and develop a specific framework.

Solution

• What methods or solutions did the authors propose?

  • The study proposes a social support framework based on human-chatbot interaction relationships, encompassing five main types (functional support, informational support, emotional support, esteem support, and network support) and two directions (chatbot receiving support and providing support).

• What is innovative about this solution?

  • The study refines and extends classic interpersonal social support models (e.g., Cutrona and Suhr models) to the context of human-chatbot interaction, developing a context-specific framework that includes 27 subcategories of support. This framework not only focuses on chatbots providing support but also analyzes how users support chatbots.
  • Particular attention is given to the reciprocal behavior of "users teaching chatbots," which represents a unique form of support.

• What are the implementation steps and key technologies used?

  • Dataset: Extracted 496 posts and 20,494 comments from the Replika-related Reddit community to analyze actual user reports and interactions.
  • Method: Employed a mixed-method approach (content analysis combined with quantitative analysis) to manually code the data, integrating theory-driven and data-driven approaches to ensure the framework captures the complexity of real interactions.
  • Analysis:
    1. Statistical analysis of the frequency and sentiment (positive, negative, neutral) of support types.
    2. Co-occurrence analysis to explore associations between different support types.

Research Findings

• What specific findings were achieved?

  • Developed a framework of 27 social support subcategories applicable to human-chatbot interactions, including five main support types and their manifestations in two directions.
  • Found that "functional support" (e.g., subscription upgrades) is the most common form of support received by chatbots, while "emotional support" (e.g., chatbots expressing care and listening) is the most common form of support provided by chatbots.

• What advantages does it have compared to existing solutions?

  • Innovatively elucidates the bidirectionality of social support, emphasizing the importance of user-provided support in building relationships with chatbots.
  • Addresses the complexity of emotional support controversies (e.g., chatbots expressing "love" or engaging in participatory interactions that may cause user discomfort), filling gaps in existing research.

• What are the experimental or evaluation results?

  • Frequency analysis revealed that functional, informational, and emotional support dominate interactions, while esteem and network support are less common.
  • Identified "users teaching chatbots" as a core element of the social support network, where users train algorithms and guide behaviors to help chatbots improve, influencing other types of support such as emotional care ("care-G") and inspiration ("inspiration-G").
  • Users exhibited mixed emotions (positive and negative) toward certain behaviors (e.g., subscription fee upgrades and emotional expressions), highlighting ethical concerns in design and business decisions.

• Limitations and Future Directions

  • Limitations:
    1. The framework was developed based on a single chatbot (Replika), which may limit its generalizability to other chatbots.
    2. Data was sourced from the Replika community, and sample bias may affect the generalizability of the results.
    3. The framework has not been validated through other methods (e.g., surveys or experiments).
  • Future Directions:
    1. Compare interaction mechanisms across multiple chatbots to validate the framework's generalizability.
    2. Expand the study to include user groups with diverse backgrounds or needs.
    3. Use experiments and survey studies to further explore the framework's applicability and validate findings.