Interrupting for Microlearning: Understanding Perceptions and Interruptibility of Proactive Conversational Microlearning Services

Document Title

Interrupting for Microlearning: Understanding Perceptions and Interruptibility of Proactive Conversational Microlearning Services

Document Information

• Subject Area: Human-Computer Interaction, Smart Home, Microlearning
• Keywords: Smart Speaker, Microlearning, Conversational Interaction, Timing Selection, Interruptibility

Research Background and Issues

• Problems or Challenges:

  1. Microlearning requires frequent engagement, but traditional learning methods are difficult to integrate into daily life.
  2. The widespread use of smart speakers in households offers new possibilities for proactive learning, but inappropriate timing for learning may disrupt users' lives.
  3. Research on timing selection for proactive learning interactions lasting more than one minute is limited, with current studies primarily focusing on short interactions (e.g., notifications within one minute).

• Significance: Proactive microlearning services can effectively utilize fragmented time to improve learning efficiency, but their interruptibility may negatively impact user experience. Exploring "appropriate interruption timing" under different time periods, locations, and activity contexts is crucial for optimizing the educational functionality of smart speakers.

• Research Motivation and Related Work:

  1. Existing studies mostly focus on short interaction tasks, with insufficient research on the interruptibility of longer microlearning sessions (3–10 minutes).
  2. Research on proactive interactions with smart speakers is largely based on hypothetical scenarios, lacking validation in real-world applications.
  3. Understanding user perceptions and appropriate timing is necessary to optimize the design of conversational microlearning services.

Solution

• Proposed Method or Solution: The authors propose a proactive conversational microlearning service (SpeechMaster) designed to randomly deliver language learning tasks via smart speakers in a household environment.

• Innovation:

  1. Investigated user perceptions and appropriate timing for proactive interactions lasting more than one minute, addressing gaps in existing research.
  2. Allowed users to dynamically control learning duration and collected reasons and contextual data upon task termination.
  3. Integrated quantitative and qualitative analysis methods to comprehensively reveal factors influencing interruptibility during microlearning.

• Implementation Steps and Key Technologies:

  1. Designed and optimized the "SpeechMaster" service through six small-scale trials involving 29 participants.
  2. Conducted a three-week study with 28 users in household settings, analyzing their learning behaviors and contextual data.
  3. Used linear mixed models to analyze the relationship between interruptibility and user activities, spatial, and temporal contexts.
  4. Proposed a five-step microlearning process: activity inquiry, availability inquiry, voice shadowing, continuous learning inquiry, and reason-for-stopping inquiry.

Research Outcomes

• Specific Findings:

  1. Proactive microlearning pushes can increase learning opportunities but require appropriate timing and context; otherwise, they reduce user engagement.
  2. Approximately 49% of cases showed users opted out of learning tasks, but proactive learning content aligned with user interests was perceived as "non-disruptive."
  3. Users' interruptibility strongly depends on their current activity type, location, and temporal context, such as:
     • Higher likelihood of engaging in longer learning sessions during low-productivity activities (e.g., resting, media consumption).
     • Specific spaces (e.g., areas where smart speakers are installed) significantly influence interruptibility due to voice recognition and sound range.

• Advantages Compared to Existing Solutions:

  1. Addressed gaps in timing research for long-duration interactions.
  2. Provided detailed contextual analyses of interruptibility, offering valuable practical guidance for application design.

• Experimental and Evaluation Results:

  1. Quantitative analysis showed the highest interruption rate (78%) for short interactions (≤1 minute), while 10-minute interactions had an interruption rate of only 25%.
  2. Questionnaire interviews revealed users preferred receiving learning tasks during "non-productive" activities, with the highest interruption rates observed during resting periods.

• Limitations and Future Directions:

  1. The study assumes that participation rates in long-term learning scenarios can infer short-term interruptibility, requiring further validation of this assumption.
  2. Current learning services monopolize smart speaker resources, and future exploration is needed to coordinate with other services.
  3. Comprehensive consideration of multi-modal interactions (e.g., smart speakers with screens) and their adaptability and impact is necessary.

Conclusion

This study, based on real-world data and contextual analysis of 28 users over three weeks, revealed factors influencing the interruptibility of long-duration microlearning tasks and proposed recommendations for optimized design and timely learning. The research expanded the educational functionality of smart speakers in household environments, demonstrating how appropriate timing can enhance user engagement and experience. It also provides theoretical support and design frameworks for future smart home educational services.