Courses
Can AI match human empathy?
Aisha Mohamed, an MSc Cyberpsychology student at ATU, is exploring how university students perceive empathy in responses generated by artificial intelligence (AI) compared with those from human therapists.
Her research project, Evaluating Perceived Empathy in AI and Human Responses to Academic Stress, examines whether differences in empathy between AI-generated and human responses are genuinely noticeable and what shapes people’s perceptions of them.
With a background in psychology, Aisha helped create Maanjiheeye, a free platform connecting university students with Somali therapists around the world. The initiative was designed to make mental health support more accessible, particularly for students displaced by conflict and dealing with academic pressures, trauma and everyday life challenges. That experience later inspired the direction of her MSc research.
“Even though the service runs 24/7, human therapists naturally aren’t available at all times,” she explained. “This led me to consider the potential of introducing a chatbot to respond to students during those gap hours when a therapist isn’t immediately online.”
As she explored the topic further, she encountered contrasting views on AI’s role in mental health support.
“The more I looked into it, the more conflicted opinions I encountered regarding AI’s role in mental health,” she said. “This made me think, ‘Why not conduct my own research to find out how students actually perceive it?’”
The study presents participants with a scenario involving a college student experiencing academic stress. Participants are shown responses without knowing whether they were written by AI or a human therapist and are asked to rate which response feels more empathetic and explain the reasons behind their choices.
“We assume the human is automatically perceived as more empathetic,” Aisha said, “but the study is really about whether that difference is actually noticeable and what drives people’s perception of it.”
The findings showed that human therapists outperformed AI in scenarios involving complex emotional distress, while AI-generated responses performed equally well in scenarios involving more straightforward academic stress.
Aisha hypothesised that participants would rate human responses as more empathetic than AI-generated responses and that factors such as AI use and previous therapy experience might influence perceptions.
Her focus on university students is intentional, as they are among the most frequent users of AI technologies, yet there is limited research on how they perceive AI in a therapeutic context. The study also builds on her longstanding interest in supporting student mental health.
According to the Higher Education Policy Institute (HEPI), 92% of university students use generative AI tools in some capacity. In Aisha’s study, 58% of participants reported using AI frequently, with 28% using it weekly and 30% using it daily.
Aisha believes the research is particularly relevant at a time of growing demand for mental health support. She notes that barriers such as cost, accessibility, availability and concerns about stigma can make traditional services difficult to access for some young people.
People are already turning to tools like ChatGPT when they are stressed or struggling, so understanding exactly how that AI-generated support is perceived really matters.
The study also explores which qualities make responses feel empathetic to students, including how emotions are acknowledged, how understanding is expressed and how personal a response feels.
Qualitative findings revealed notable differences between groups. Students with previous therapy experience were more likely to value active listening and responses that focused on exploring and understanding a problem before offering solutions. Students who frequently used AI tools, meanwhile, responded positively to practical, action-oriented guidance and viewed clear, step-by-step solutions as supportive and empathetic.
“That’s something that could actually be used to improve how digital mental health services are built,” she added.
Looking to the future, Aisha believes AI could potentially complement traditional mental health services but says careful regulation and safeguards will be essential.
“AI is getting really good at mimicking human empathy, and many people are already using it for emotional support,” she said. “Under very strict and regulated conditions, AI could definitely be a useful tool to complement traditional mental health services.”
At the same time, she cautions that there are risks if AI systems are used without appropriate guardrails.
“There is a risk it could cause more harm than benefits, like emotional dependency,” she explained. “Traditional therapy has distinct stages of personal growth, including a termination phase, whereas commercial AI is often designed to keep users constantly engaged.”
Photo caption: Aisha Mohamed, an MSc Cyberpsychology student at Atlantic Technological University
