MeetingCoach: An Intelligent Dashboard for Supporting Effective & Inclusive Meetings

Title of the Paper

MeetingCoach: An Intelligent Dashboard for Supporting Effective & Inclusive Meetings

Paper Information

• Subject Area: Human-Computer Interaction, Video Conferencing, AI-driven Meeting Feedback Tools
• Keywords: meetings, feedback, team dynamics, inclusivity, video conferencing, behavior monitoring, artificial intelligence, emotion recognition, user interface, design recommendations

Research Background and Problem Statement

• Identified Problems and Challenges:

  • Group dynamics in meetings (e.g., speaking time, participation) can lead to efficiency and inclusivity issues.
  • The COVID-19 pandemic in 2020 forced a shift to video conferencing, which struggles to capture non-verbal social signals, affecting meeting outcomes.
  • The fragmented ecosystem of video conferencing exacerbates the marginalization of certain group members during meetings.
  • Current video conferencing platforms lack effective post-meeting feedback mechanisms to improve meeting efficiency and inclusivity.

• Importance of the Problem:

  • For companies, improving meeting efficiency and inclusivity can yield significant economic benefits and enhance employee job satisfaction.
  • Providing technological tools to help users understand meeting behavior dynamics is a crucial direction for improving meeting experiences.

• Research Motivation and Related Work:

  • Previous studies show that balanced team participation and positive interaction of non-verbal signals significantly impact team performance, but these dynamics are difficult to replicate in video conferencing.
  • While some real-time feedback mechanisms or emotion visualization systems, such as EMODASH, exist, they often lack post-meeting data analysis and actionable feedback suggestions.

Solution

• Proposed Method or Solution:

  • Develop an intelligent post-meeting feedback dashboard called MeetingCoach, designed to provide post-meeting feedback through behavioral and content analysis, helping users understand meeting dynamics.
  • The dashboard integrates AI technologies to extract meeting features and presents quantified information on participation, emotions, issues, consensus, and more.
  • Offers a personalized feedback experience, including temporal visualization of behavioral data and actionable suggestions.

• Innovative Features:

  • Combines bimodal data analysis of emotional and behavioral signals.
  • Provides both summary and timeline visualizations of feedback, along with specific actionable suggestions.
  • Includes privacy protection features based on user needs, allowing flexible data-sharing settings (e.g., making emotional data private and accessible only to the individual).

• Implementation Steps and Key Technologies:

  • Conducted a two-phase design process: first, a needs analysis survey to identify key meeting feedback requirements; second, a long-term user study to record meeting data and develop a dashboard prototype.
  • Used Microsoft Teams platform and custom recording bots to capture audio and video data.
  • Developed multimodal sensing algorithms, including:
    • Transcription technology to extract text and identify issues.
    • Facial expression recognition to classify emotions and quantify positive, neutral, and negative emotions.
    • Head movement detection (nodding and shaking) to identify consensus signals.
    • An interactive, personalized dashboard built using HTML and D3.js for the user interface.
  • Conducted multiple design iterations, collecting user feedback through interviews and surveys to refine the dashboard.

Research Outcomes

• Specific Outcomes:

  • Developed the MeetingCoach system prototype, capable of analyzing and presenting meeting dynamics and improving users' understanding of meeting behaviors through post-meeting feedback.
  • Feedback provided covers multiple modules, including participation time, speaking order, consensus events, emotional modulation, and meeting issues.

• Advantages Over Existing Solutions:

  • Compared to traditional meeting notes, the system more effectively identifies group dynamics and provides suggestions to optimize future meetings.
  • Less disruptive than real-time feedback tools and more comprehensive and accurate than emotion recognition methods limited to a single modality.

• Experimental or Evaluation Results:

  • The system significantly improved users' perceptions of meeting efficiency and inclusivity during experiments (effectiveness score M=4.45, inclusivity score M=5.23).
  • Users found the video and speaking order features particularly useful but expressed some uncertainty about the application scenarios for emotion-related features.
  • Users generally supported actionable suggestions based on behavioral history and the privacy settings of the personalized dashboard structure.

• Limitations and Future Directions:

  • Meeting data was collected from internal teams within a single company, and external generalizability needs further validation.
  • The current system primarily targets remote meeting scenarios; future exploration is needed to support hybrid meetings.
  • Most experimental participants used the dashboard only once, leaving the long-term impact unverified.
  • Future research could explore how AI systems can better understand meeting contexts and develop behavior training features based on long-term data.

Conclusion

MeetingCoach provides an innovative post-meeting feedback dashboard solution, improving users' focus on remote meeting efficiency and inclusivity through behavioral timelines and actionable suggestions. Future research should expand the system's applicability while continuously optimizing user experience and data privacy protection features in practice.