Empathy in leadership communication: Experimental evidence from two vignette studies on AI's role in message improvement and observers' perceptions
Didem Sedefoglu-Ulucak, Sandra Ohly, Joachim Schmelz · Computers in Human Behavior · 2026
The increasing availability of generative AI tools raises the question of whether they can meaningfully support leadership communication. One use case discussed in the literature is AI’s potential to enhance empathetic communication. Yet little is known about whether leaders are willing to use AI’s support or how observers interpret such communication. This research examines the topic from two complementary perspectives using two vignette-based experiments with working adults in Germany. In Study 1, participants in the leadership scenario received advice framed as coming from an AI system or a human communication specialist before revising a previously written feedback email. We compared empathy levels of emails before and after giving advice to determine whether participants were willing to use AI advice. The likelihood of revising the email in response to the advice did not differ depending on the human vs. AI condition, nor did we find differences in terms of improvement of empathy. Study 2 investigated how observers interpreted leadership communication with varying degrees of AI involvement (no AI, low AI, and high AI). High AI use (generating the whole message with AI) led to lower evaluations of empathy, credibility, and feedback quality compared to the no-AI condition, whereas low AI use (using AI to fix wording) did not result in negative reactions. Tentative evidence suggests that AI literacy may mitigate the negative effect of high AI involvement on perceived empathy, but it did not moderate the effects on credibility or quality. Together, the findings show that while individuals were open to integrating AI advice when communicating in a leadership scenario, high AI involvement can negatively affect how observers interpret their messages.