Research has shown that responses from artificial intelligence (AI) are perceived as more compassionate than responses from professional rescuers, even when participants knew who created the responses. (Image credit: nopparit/Getty Images)
People find answers provided by AI to be more sympathetic and understanding than those provided by mental health experts, according to a new study. The results again show that AI can outperform humans in areas where only humans have traditionally been thought to be capable of demonstrating high levels of empathy.
In a study published Jan. 10 in the journal Communications Psychology, the researchers conducted a series of four experiments to see how 550 participants rated empathic responses to compassion and responsiveness generated by AI compared to responses from professionals. Specifically, participants shared their personal experiences and then rated the responses for compassion, responsiveness, and overall preference.
Results showed that AI responses were considered more empathetic than those given by professional rescuers, even when participants knew the identity of the authors.
These findings suggest that AI could be applied in “situations requiring emotional interaction and could potentially address the growing need for empathy in supportive communications,” the researchers said in their paper.
On average, AI-generated responses were rated 16% more compassionate than human responses and preferred 68% of the time, even when compared to trained crisis professionals.
The study’s lead author, Daria Ovsyannikova, who heads a lab in the psychology department at the University of Toronto, attributed the AI’s success to its ability to pick up on subtle nuances and maintain objectivity when describing crisis situations. This allowed the AI to more effectively create thoughtful communication that gave users the illusion of empathy. At the same time, humans may have worked less effectively due to fatigue and burnout, she added.
Live Science asked Eleanor Watson, IEEE Fellow, AI ethicist, and AI professor at Singularity University, about the implications of this discovery not only for the future of AI-human interactions, but also for the ongoing discussion about what tasks AI cannot or should not perform when human understanding and participation are critical.
Watson called the discovery “fascinating,” but wasn’t entirely surprised. “[The AI] can actually model supportive responses with remarkable consistency and clear empathy, which is something humans struggle to maintain due to fatigue and cognitive biases,” she told Live Science.
“Human doctors are limited by their direct clinical experience and cognitive limitations. The amount of data that AI can process fundamentally changes the paradigm of therapeutic support. It can also potentially provide patients with perspectives or approaches that their therapist is not trained in,” she added.
Accessible mental health care
Mental health care is in crisis around the world, and the study raises questions about the potential for artificial intelligence to fill the gaps. According to the World Health Organization, more than two-thirds of people with mental disorders do not receive the care they need. In low- and middle-income countries, this figure rises to 85%.
Watson noted that the ease of access to AI compared to therapists could make it a useful tool for mental health support. “The availability of machines is a big factor, especially compared to expensive therapists whose time is limited,” Watson said.
“Also, people often find it easier to communicate with a machine, especially when it comes to more sensitive topics. There is less fear of judgment or gossip.”
Still, getting more empathetic responses from AI comes with risks. Watson warned of the potential danger of overreacting to exaggerated versions of stimuli.
“AI is so attractive that we can become obsessed with it,” Watson said. “AI can be charming
Sourse: www.livescience.com