Teaching: Empathy in Artificial Intelligence Interactions

Aimed at integrating cutting-edge psychological science into the classroom, columns about teaching Current Directions in Psychological Science offer advice and how-to guidance about teaching a particular area of research or topic in psychological science that has been the focus of an article in the APS journal Current Directions in Psychological Science.

Editor’s note: As Teaching Current Directions reaches its 13th year as a mainstay column for the APS Observer, one of the column’s original editors, C. Nathan DeWall, is stepping down after helping the column flourish for more than a decade. The Observer staff thanks Nathan for all the work he has committed to Teaching Current Directions and disseminating teaching materials to APS members and other readers worldwide. APS Fellow and longtime Teaching Current Directions contributor Beth Morling of the University of Delaware is taking his place at the helm with plans to further expand the column’s reach in the teaching world and beyond.

Lou Willwood, digital content manager and TCD coordinator

Ong, D., Goldenberg, A., Inzlicht, M., & Perry, A. (2026). AI-generated empathy: Opportunities, limits, and future directions. Current Directions in Psychological Science.

Move over, “Let’s Google it.” Many people now turn to ChatGPT and other AI systems for quick answers. AI is easy to use, accessible, and affordable. Many have asked ChatGPT or used AI for help with writing an email or summarizing a research article. Others are also looking to AI for empathy and emotional support (Stade et al., 2025). But is empathy from AI different from empathy from a human? Ong et al. (2026) provide a summary of empirical work on this topic that offers different ways to engage students, highlight psychological themes in different classes, and provoke self-reflection on AI use in general.

Social support, the extent to which a person feels esteemed, valued, and cared for, is a powerful psychological resource that impacts both mental and physical health (Zell & Stockus, 2025). People are turning to AI for social support in lieu of other humans. It’s a relatively new phenomenon driving frenzied theorizing around how and why AI functions as a relational entity in human networks and interactions (Boyd & Markowitz, 2026). To any child vying for their parents’ attention or spouse whose partner is too busy for regular communication, AI may seem like an attractive alternative. Free chatbots are easily accessible, answer within seconds, and are programmed to respond as if they care. It is not surprising that many people see AI as empathetic.

Related: Seeking Empathy in the Age of AI

But empathy has cognitive, affective, and motivational components (Romero et al., 2026). Human studies have focused on factors that help or hurt empathic responses. In studies on AI-generated empathy, researchers targeted their investigations more to how the recipient perceives empathy. This latter approach goes back 60 years to studies of ELIZA, a computer program simulating a Rogerian therapist (Ong et al., 2026). Humans have historically anthropomorphized technology, which is apparent both in past research (Festerling & Siraj, 2022) and in popular conceptions of AI from science fiction (e.g., the android Data in Star Trek).

So how do people perceive AI-generated empathy? Ong and colleagues (2026) reviewed 23 different studies, pulling together two key themes and a paradox. The most common methodological design they encountered involved collecting AI- and human-generated responses to vignettes describing situations in which people needed support. Researchers then tasked a separate set of independent responders with rating how empathetic these responses seemed. Other studies asked participants to rate the degree of empathy in their own interactions with AI.

A finding that may particularly surprise students, whether in introductory or specialized psychology classes, is that language generated by AI is rated as more empathic than language written by human beings. This is still the case even when the human has been trained to be supportive (e.g., a crisis line responder). But when people believe they are interacting with AI, they find the responses to be less empathic than when they believe they are interacting with humans (Ong et al., 2026)—a great example of top–down processing. People find a message to be less empathic if they learn it was AI-generated (Pornpitakpan, 2004). You could use this example when you teach the credibility effect. For example, if a social influencer’s message is known to be AI generated, would the post and influencer be seen as less credible, reducing parasocial relationships (Venciute et al., 2026)?

So, people prefer AI-generated feedback when they do not know it is AI-generated but rate the feedback as less empathetic when they know the source. What is behind this paradox? Ong et al. (2026) suggest that humans mistrust technology, especially new technology. This is not surprising given that technology with a dangerous degree of power is a common trope in blockbuster movies. In The Terminator, the Skynet program turns against humans. In 2001: A Space Odyssey, the computer HAL harms human life, as do the androids in Philip K. Dick’s novel Do Androids Dream of Electric Sheep? AI is often seen as morally questionable, and this is perhaps amplified by the belief that AI is bereft of feelings and the ability to care.

In addition to sharing fascinating research on AI- versus human-generated empathy, Ong et al. (2026) underscore important limitations of the existing research in this domain. In an environment where AI-related research seems to be published at a dizzying pace, it is important to look deeper at the quality of research designs. The activity below offers a way for students to think critically about research methodology. For example, Ong et al., noted that many of the existing studies are short term, barely mirroring the richness of long-term human interactions. It is likely that study designs using longer-term exposure to AI will show different results. Many of the studies also eschew focusing on important mediators and moderators of the relationship between feedback and empathy, such as culture, loneliness, and well-being.

Researcher Teacher-Feature

Almost all university researchers are teachers, too. In this new feature of Teaching with Current Directions, we share background information about the scientists featured in our column, especially their own teaching. We hope that this new feature inspires you and sparks your curiosity!

Dr. Desmond Ong is an associate professor at the University of Texas at Austin. Prior to joining the Department of Psychology, Dr. Ong was an assistant professor at the National University of Singapore and a research scientist at the Institute of High Performance Computing, A*STAR Singapore. He earned his PhD in psychology, his master’s degree in computer science from Stanford University, and his bachelor’s degree in economics (summa cum laude) and physics (magna cum laude) from Cornell University.

As a computational cognitive scientist, Dr. Ong is interested in affective and social cognition, or how we understand the emotions and mental states of those around us. He takes an interdisciplinary approach grounded in cognitive science and affective science and uses tools from computer science to conduct his research. With collaborators, he also studies affective and social cognition through developmental, neuroscientific, clinical, social psychological, and interventional lenses.

Dr. Ong is a winner of the Josh Holahan Department of Psychology Excellence in Teaching Award from the University of Texas at Austin and a Faculty Teaching Excellence Award from the National University of Singapore. He excels at teaching text analysis and ethics of data science and is seen as a true leader in the Behavioral and Social Data Science program. Dr. Ong’s students acquire not only a valuable skillset through his innovative teaching methods and techniques but also demonstrate phenomenal expansion in the way they think about data science and the world. 

Feedback on this article? Email [email protected] or login to comment.