Written by: Bob Bertsch
A recent study (Yin et al., 2024) found that in an online setting artificial intelligence (AI) was able to generate responses that made human recipients feel more heard than responses generated by other humans. While the results of the study don’t suggest AI can replace humans in making people feel heard, they do point to some important lessons we can learn from AI.
Yin, Jia, and Waksiak (2024), asked study participants to describe a complex situation they were dealing with and indicate how much they were feeling each of six emotions (happiness, sadness, fear, anger, surprise, and disgust) as a result of the situation. The participants’ descriptions were randomly selected to be shared with either a human respondent or Bing Chat, an AI-powered assistant now known as Copilot. Both the human respondents and Bing Chat were also asked to indicate how much they thought the original participant felt each of the same six emotions mentioned above. Finally, the original participants were sent the responses to their complex situation and surveyed to measure their perceptions, feelings, and emotions after reading the responses. To assess the effect of the source of the response on the participant’s perceptions, only some of the participants were told the response they received was generated by Bing Chat.
AI Made People Feel More Heard
The results of the study showed that overall the AI-generated responses elicited more positive reactions than the human responses. Researchers found AI responses were rated higher in making participants feel heard, in the participant’s perception of the accuracy of the response, in making participants feel they were understood, and in making the participant feel connected to the responder.
The AI-generated responses were found to contain more emotional support and focused on making others feel better. The human-generated responses contained more practical support, focused on helping others solve a problem. Human responders were more likely to share their personal experiences related to the participant’s situation and provide more of their own insights. In contrast, Bing Chat included more acknowledgment of the participant’s feelings.
AI-generated responses were found to more often include techniques associated with active listening and partner responsiveness. Those techniques include:
- Repeat back key phrases/summarize the story in their own words
- Voice understanding (e.g., ‘‘I understand,’’ ‘‘I see’’)
- Acknowledge the discloser’s feelings/indicate that the feelings are justified
- Express caring for the discloser (‘‘I care about you’’)
- Express emotions or empathy (‘‘I’m sorry that happened’’; ‘‘I’m happy for you’’)
What AI Can Teach Us
It’s important to note that the human respondents in this study did not know the participants they were responding to and they had no specific training in how to make people feel heard. There’s no indication AI can make a person feel heard better than a close friend, family member, or trained professional can. However, people seeking the connection, empathy, and validation that comes from feeling heard may not always have access to someone close to them or to a professional and may choose to share their experiences online. In that scenario, AI has a lot to teach us, like favoring emotional support over practical support, and focusing more on the experience of others than on ourselves.
In a review of the literature (2023), Roos, Postmes, and Koudenburg identified five components of feeling heard. “At the interpersonal level,” they write, “people feel heard when they have 1) voice, and receive 2) attention, 3) empathy, 4) respect. At the collective level people should experience 5) common ground.” Offering attention, empathy, and respect in our responses to others online and in real life could be a key step in experiencing common ground.
References
Roos, C. A., Postmes, T., & Koudenburg, N. (2023). Feeling heard: Operationalizing a key concept for social relations. PLOS ONE, 18(11), e0292865. https://doi.org/10.1371/journal.pone.0292865
Yin, Y., Jia, N., & Wakslak, C. J. (2024). AI can help people feel heard, but an AI label diminishes this impact. Proceedings of the National Academy of Sciences, 121(14), e2319112121. https://doi.org/10.1073/pnas.2319112121
Photo by Google DeepMind from Pexels.
