Human vs. AI Empathy Perception

AI Empathy: Human vs. Machine

Large language models and AI show significant social–emotional skills, potentially enhancing human–AI interactions and AI's capacity for emotional support. It's not clear if empathy (understanding, ‘feeling with’ and caring) is perceived differently depending on whether it comes from AI or humans.

Study Overview

Nine studies (n= 6,282) were conducted where AI-generated empathic responses to participants’ emotional situations were labelled as either human or AI provided. Responses attributed to humans were considered more empathic and supportive. They also elicited more positive and fewer negative emotions compared to those attributed to AI.

Key Findings

The belief that AI assisted in generating human-attributed responses reduced the perceived empathy and support. These effects were consistent across different response lengths, delays, iterations, and large language models. The effects were mainly driven by responses focused on emotional sharing and care. People preferred human interaction over AI for emotional engagement.

These results improve the understanding of empathy and human–AI empathic interactions.

All preprocessed data and analysis files are available via OSF.