Toward Affective Empathy via Personalized Analogy Generation: A Case Study on Microaggression

Research Background and Problem

• Identified Issues or Challenges: Although the importance of empathy in diverse societies is widely recognized, existing technologies (such as immersive virtual reality, role-playing, and narrative tools) aim to evoke empathy by simulating others' experiences. However, these methods assume that identical experiences can elicit identical emotional responses, an assumption that proves significantly limited when addressing issues like microaggressions. Microaggressions are subtle and ambiguous discriminatory behaviors in daily life, and recipients' emotional responses often vary greatly due to individual differences. This variability makes it challenging to foster cognitive or affective empathy between victims and bystanders.
• Importance of the Problem: Individuals may struggle to empathize with others facing microaggressions, leading to misunderstandings, conflicts, and communication barriers—contrary to the societal coexistence demands in increasingly complex multicultural contexts. Existing technologies have not fully addressed the empathy barriers caused by differences in emotional responses among individuals.
• Research Motivation: The goal is to develop a new method that better bridges individual emotional differences, thereby enhancing the emotional empathy needed when confronting complex situations like microaggressions.

Solution

• Proposed Method: This paper proposes and develops an agent system based on large language models (LLMs) called "EmoSync." It provides users with a tool that generates customized analogous microaggression scenarios to evoke emotional resonance similar to that of the target individual.
• Innovations:
  • Utilizing LLMs to generate personalized analogous scenarios rather than merely presenting others' original experiences.
  • Bridging emotional differences through analogy, emphasizing shared emotions rather than identical experiences.
  • Focusing specifically on microaggressions, a domain characterized by strong subjective perceptions.
• Implementation Steps and Techniques:
  1. Personalized Emotional Understanding (Phase 1): Using surveys and LLM prompt engineering, the model learns a specific user's emotional response patterns to different microaggression scenarios.
  2. Analogous Scenario Generation (Phase 2): Based on the user's personalized data and the target individual's emotions, the LLM generates analogous microaggression scenarios designed to elicit emotional responses similar to those of the target individual.
  3. End-to-End Evaluation (Phase 3): Testing users to evaluate the system's effectiveness in enhancing empathy, followed by quantitative and qualitative analysis.

Research Outcomes

• Specific Achievements:
  • EmoSync effectively generates personalized analogous scenarios, enabling users to exhibit emotional responses closer to those of the target individual in specific microaggression contexts.
  • Users experienced increased cognitive empathy and affective empathy scores after using EmoSync, with significant improvements in fantasy (FS) and empathic concern (EC) metrics.
  • EmoSync demonstrated potential in helping users uncover implicit information and meanings within original microaggression scenarios.
• Advantages Over Existing Solutions:
  • Overcomes the emotional asymmetry issues caused by placing users in "identical scenarios" in existing systems.
  • Emphasizes guiding emotional resonance and cognition through familiar analogous contexts, making it more adaptable to cross-cultural scenarios.
• Experimental or Evaluation Results:
  • In the experiment, 60 participants experienced EmoSync. Emotional distance to the original microaggression scenario was reduced by approximately threefold, indicating the effectiveness of analogous scenarios in connecting users with the target individual's emotions.
  • User feedback on the "personalized resonance" of the analogous scenarios and their "helpfulness for understanding the original scenario" was predominantly positive.
• Limitations and Future Directions:
  • Limitations:
    • While analogies were effective, some generated content lacked logical or emotional relevance and could even provoke discomfort.
    • The system currently generates predominantly text-based, overly simplistic scenarios, with a lack of diverse scenario implementations (e.g., multimodal extensions).
  • Future Directions:
    • Develop more robust and diverse models to avoid erroneous analogies and expand domain applications (e.g., intergenerational communication or cross-cultural contexts).
    • Integrate the system into real-time communication or social platforms to provide instant empathy triggers.
    • Further optimize the design for emotional data collection and privacy protection.

EmoSync demonstrates the potential of generating personalized analogous scenarios to aid in empathy building, expanding the understanding of empathy systems in the HCI field in multiple dimensions. However, balancing experiential depth, emotional burden, and privacy protection in practical applications will be a key challenge for future research.