IntroBot: Exploring the Use of Chatbot-assisted Familiarization in Online Collaborative Groups

Document Title

IntroBot: Exploring the Use of Chatbot-assisted Familiarization in Online Collaborative Groups

Document Information

• Subject Areas: Human-Computer Interaction, Computer-Supported Collaboration, Artificial Intelligence and Online Team Familiarization
• Keywords: Chatbot, Online Collaboration, Team Familiarization, AI-mediated Communication, Trust and Cohesion, Topic Recommendation, Data-driven Design

Research Background and Issues

• Problems or Challenges:

  • Building trust and team cohesion among unfamiliar team members in online collaboration is crucial but difficult to achieve quickly through computer-mediated communication.
  • Conventional team-building strategies require human facilitation, which is hard to scale for large-scale online ad hoc teams.

• Significance:

  • Trust and cohesion within teams significantly impact team performance, such as task efficiency and innovation capability.
  • As remote collaboration and online team trends rise, this issue will persist in the long term.

• Research Motivation:

  • Studies indicate that familiarization among team members helps improve trust and interaction quality.
  • Exploring the potential of chatbots to support the team familiarization process, particularly their role in fostering trust and cohesion.

Solution

• Proposed Method:

  • Design and implement a chatbot named "IntroBot" to facilitate rapid familiarization among online teams.
  • Utilize social media data (e.g., Instagram) to recommend shared-interest topics and enhance interaction quality through photo sharing and interaction management.
  • Incorporate three functional aspects: Knowledge Support (topic recommendation, photo sharing), Management Support (time management, topic guidance), Social Support (promoting trust and interaction).

• Innovations:

  • Automatically generate topic recommendations using users' social media data, overcoming the limitations of human facilitation.
  • Provide dynamic interaction management features, such as detecting and reviving interrupted conversations.
  • Support personalized topic recommendations and foster trust and emotional connections among users through content-based (photo) interactions.

• Implementation Steps and Core Technologies:

  1. Extract keywords from users' social media (Instagram), calculate keyword similarity to generate shared-interest topics.
  2. Apply natural language processing techniques (e.g., word embeddings and similarity calculations) for personalized topic recommendations.
  3. Offer photo-sharing functionality, with user permission mechanisms to mitigate privacy concerns.
  4. Implement conversation management features, including dynamic correction of conversation interruptions, time reminders, and information guidance.
  5. Develop an iOS application to enable real-time chat functionality, utilizing NoSQL databases to log conversations for scalability.

Research Outcomes

• Specific Results:

  1. Groups assisted by IntroBot showed significantly higher levels of trust, cohesion, and interaction quality compared to control groups.
  2. Participants using IntroBot generated approximately 1.8 times more creative outputs (task performance) than those in free conversation groups.
  3. Through photo sharing and topic recommendations, participants became familiar with each other more quickly and established higher levels of social connection.

• Advantages:

  1. IntroBot surpasses traditional human-facilitated familiarization processes, requiring no human intervention and being easily scalable.
  2. Automated mechanisms significantly reduce the awkwardness of initial conversations, enhancing efficiency.
  3. Provides data-driven social support, not limited to topic selection but also fostering deep visual and emotional engagement.

• Experiments and Evaluation:

  • Compared the performance of 60 participants under three experimental conditions (IntroBot-assisted, free conversation, no conversation) to validate the chatbot's significant positive impact on trust, cohesion, interaction quality, and team performance.
  • Qualitative analysis revealed that photo sharing as social cues helped users quickly establish intimacy and trust.

• Limitations and Future Directions:

  • Limitations:
    • The study primarily involved young, Korean-speaking users, which may limit generalizability.
    • The experimental setup was simplified (e.g., two-person teams, single tasks), requiring validation in more complex scenarios.
    • Privacy concerns and risks of inappropriate topic recommendations need ongoing resolution.
  • Future Directions:
    • Expand to other social media data sources to enhance topic diversity.
    • Explore applications in multi-member teams and diverse tasks (e.g., long-term collaboration, complex planning).
    • Optimize privacy protection technologies, such as sensitive content filtering and automated violation detection.

Conclusion

This study introduces an innovative chatbot design for team familiarization and provides empirical evidence of its effectiveness in enhancing team collaboration efficiency. It offers significant insights into human-AI interaction design within online collaboration environments.