Summary: Researchers have created an AI system called EmoSync that enhances empathy by generating emotional analogies tailored to each user’s personality, values, and life experiences. Rather than assuming a single emotional response fits everyone, EmoSync uses a large language model (LLM) to map individual traits and craft personalized scenarios that make others’ feelings more relatable and understandable.
In a study involving more than 100 participants from varied backgrounds, people who used EmoSync showed notably better understanding of others’ emotions than those relying on conventional empathy aids. This work marks an important step in emotion-aware AI, offering practical ways to promote authentic interpersonal understanding across diverse social settings.
Key Facts:
- Personalized empathy: EmoSync leverages LLMs to produce tailored emotional analogies that reflect each user’s personality, experiences, and values.
- Validated outcomes: Participants using EmoSync demonstrated statistically meaningful improvement in recognizing and understanding others’ emotions compared with traditional approaches.
- Real-world relevance: By translating another person’s emotional state into situations the user can personally relate to, EmoSync helps bridge emotional gaps across different identities and life histories.
Source: POSTECH (Pohang University of Science and Technology)
Overview of the research
A research team at POSTECH in South Korea developed EmoSync, an AI-driven method intended to deepen emotional understanding between people. The system analyzes an individual’s psychological profile and known emotional response patterns, then uses a large language model to generate metaphorical or analogical scenarios that connect another person’s feelings to situations familiar to the user. This personalization acknowledges that people respond differently to similar events depending on their history, temperament, and values.
Modern societies are composed of individuals with widely varying backgrounds and perspectives. Even when two people face the same event, their emotional responses can diverge sharply. Conventional computer-mediated empathy tools often assume uniform emotional mappings — that presenting the same scenario will produce similar feelings for everyone. EmoSync challenges that assumption by explicitly modeling and leveraging individual differences to produce analogies that resonate personally with each user.
For instance, when a user has difficulty empathizing with a colleague who experiences subtle workplace exclusion, the system examines the user’s own life experiences and suggests a parallel that is emotionally meaningful to that specific person — such as recalling a time of exclusion during school — so the emotional content becomes concrete and relatable. By reframing another person’s feelings through the user’s own emotional frames of reference, EmoSync makes abstract emotional states more accessible.
The research team reported a controlled study with over 100 participants from diverse demographic and experiential backgrounds. Compared to traditional empathy training or static scenarios, participants who interacted with the personalized analogies produced by EmoSync showed significant gains in their ability to recognize and understand others’ emotions. The findings provide empirical support for the idea that customized metaphorical experiences can meaningfully increase empathy in practice.
Hyojin Ju, the study’s first author, noted that this work highlights how AI can facilitate genuine understanding and empathy between people. The team intends to continue refining the approach and developing AI tools that support empathetic engagement in real-world social situations. Professor Inseok Hwang added that the project demonstrates generative AI’s capacity to identify unique emotional structures and create personalized experiences that reliably evoke specific empathic responses, representing both academic and practical advances in the field.
The research was carried out by Professor Inseok Hwang and Ph.D. students Hyojin Ju, Jungeun Lee, and Seungwon Yang in the Department of Computer Science and Engineering at POSTECH, in collaboration with Professor Jungseul Ok.
Funding: The project received support from the National Research Foundation of Korea (NRF) Mid-career Researcher Program, the Future Convergence Technology Pioneer Project funded by the Korean government (MSIT), and the University ICT Research Center Project from the Institute of Information & Communications Technology Planning & Evaluation (IITP), also funded by MSIT.
About this AI and empathy research news
Author: Jinyoung Huh
Source: POSTECH, Pohang University of Science and Technology
Contact: Jinyoung Huh – POSTECH
Image credit: Neuroscience News
Original research presentation: The research findings were presented at ACM CHI 2025, where the work received the “Popular Choice Honorable Mention Award,” placing it among the top 5% of 74 demonstrations in the Interactivity track.