Crafty Compassion: linguistic and discursive dimensions of empathy in AI communication

Crafty Compassion: Linguistic and discursive dimensions of empathy in AI Communication Vivian M. De La Cruz Dept. of Humanities, Languages and Education University of Enna “Kore” vivian.delacruz@unikore.it Abstract In humans, empathy is a multifaceted phenomenon that necessitates the application of both cognitive and emotional intelligence. With the advent of advanced artificial intelligence, particularly large language models (LLMs) like ChatGPT-4, there has been a significant shift in how AI systems interact with humans, and a crucial aspect of these interactions is the capacity for AI to convey (or mimic) empathy through language, a feature that holds profound implications for user experience and trust (Liu & Sundar, 2018; Shteynberg et al. 2024). This talk will reflect on the linguistic and discursive dimensions of empathic AI, exploring how the use of empathic language by LLMs influences human perceptions of empathy. Through an interdisciplinary approach that draws from cognitive science, discourse- pragmatic analysis, and AI research, this study examines the linguistic features and discursive strategies that LLMs employ to generate a sense of empathy, compared to those used by humans with other humans (Concannon et al. 2023; Lee et al. 2024). Key questions addressed include: What are the linguistic markers of empathy in AI-generated text? How do these markers shape user perceptions and interactions? What role does context play in the effectiveness of AI-generated empathic language? By analyzing a number of case studies, this research aims to further the understanding of how language, and empathic language in particular, is increasingly being used to mediate the relationship between humans and AI, contributing to the broader discourse on AI and human communication (Inzlicht et al. 2023, Kidder et al. 2024).