Paper Type
Short
Paper Number
1133
Description
Large language models (LLMs) based on generative AI have significantly impacted human society, while still face issues like inaccurate responses, slow performance, and lack of emotion. To address these concerns, this paper collects user reviews from app stores in 21 countries and 10 languages, employs the LDA model to extract factors and explores the gap between user ratings and textual emotions as indicators of forgiveness and perceived bottlenecks. Sentiment analysis is conducted using the Word2vec-SVM model, followed by empathy-based attributions. The findings reveal that functional and economic remedies effectively evoke empathy and forgiveness, while empathic solutions successfully overcome bottlenecks. Interestingly, empathetic users tend to be more discerning. Further SNA uncovers that timely problem-solving, software flexibility, model-updating, voice-image analysis contribute to breaking bottlenecks. Additionally, heterogeneity analysis demonstrates that Eastern users exhibit greater sensitivity towards price and are more likely to forgive through economic remedies; whereas Western users prioritize app quality improvements.
Recommended Citation
Yang, Yixing; Huang, Jianxiong; and Guo, Xuefei, "Viva Understanding or Narrowly Okay? The Impact of Empathy on User Forgiveness and Perceived Bottlenecks in LLMs" (2024). PACIS 2024 Proceedings. 5.
https://aisel.aisnet.org/pacis2024/track01_aibussoc/track01_aibussoc/5
Viva Understanding or Narrowly Okay? The Impact of Empathy on User Forgiveness and Perceived Bottlenecks in LLMs
Large language models (LLMs) based on generative AI have significantly impacted human society, while still face issues like inaccurate responses, slow performance, and lack of emotion. To address these concerns, this paper collects user reviews from app stores in 21 countries and 10 languages, employs the LDA model to extract factors and explores the gap between user ratings and textual emotions as indicators of forgiveness and perceived bottlenecks. Sentiment analysis is conducted using the Word2vec-SVM model, followed by empathy-based attributions. The findings reveal that functional and economic remedies effectively evoke empathy and forgiveness, while empathic solutions successfully overcome bottlenecks. Interestingly, empathetic users tend to be more discerning. Further SNA uncovers that timely problem-solving, software flexibility, model-updating, voice-image analysis contribute to breaking bottlenecks. Additionally, heterogeneity analysis demonstrates that Eastern users exhibit greater sensitivity towards price and are more likely to forgive through economic remedies; whereas Western users prioritize app quality improvements.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
AI