Artificial Intelligence
Digital Empathy May Outperform Humans
Can technology provide the warmth and understanding we (seemingly) crave?
Posted January 3, 2025 Reviewed by Tyler Woods
Key points
- Empathy may enhance mental health treatment, but does AI offer the same benefit?
- Research participants initially doubted the effectiveness of AI, but ultimately benefited from it.
- Digital empathy led to swift, lasting reductions in seven negative emotions.
- Improvements are attributed in part to user expectations, indicating empathy is one piece of a larger puzzle.
Empathy is considered by many to be a cornerstone of effective mental health treatment. American psychologist Carl Rogers achieved fame with his claim that empathy was the “necessary and sufficient condition” for change in therapy, and many experts agree that empathy creates feelings of trust, reduces negative feelings, and boosts self-esteem.
Can digital devices do the same thing?
Digital mental health apps are enjoying a true breakthrough moment, and AI shows great promise in mimicking human conversation. But can digital empathy compete with human empathy in terms of healing?
And for that matter, how important is empathy to begin with, and does it even matter where it comes from?
Digital Empathy Outperforms Expectations
As part of an ongoing venture, I recently had the opportunity to put these questions to the test by measuring the impact of computer-generated responses in a newly designed application on individuals seeking relief from negative thoughts and feelings.
A team of developers and I questioned—and sought to measure in a meaningful way—whether people would benefit from so-called “digital empathy.” After all, a computer can’t feel anything, much less empathy for a human.
As it turns out, it doesn’t appear to matter.
We surveyed 290 beta testers who used the app in the winter of 2023-24. At the initial evaluation, we asked them to rate the level of warmth and understanding they experienced with friends and loved ones on a scale from 0 (not at all) to 100 (perfect). Then, we asked them to predict how much warmth and understanding they’d receive from the app.
Sadly, most beta testers didn’t report much warmth and empathy from friends and loved ones, and they didn’t expect much warmth and understanding from the app, either. Perhaps this reflected the long-held belief that technology is inherently impersonal and “less than” human.
Participants then used the app, which was designed to respond just as I would if I were meeting with them 1-on-1. The app expressed understanding, offered warmth, and proposed many new ways of challenging the distorted thoughts that triggered their negative feelings.
Once they began interacting with the app, their perceptions of digital empathy changed dramatically. In fact, after three days, and at the end of the four-week beta test, they rated the digital warmth and understanding nearly twice as high as the human variety and far higher than they had predicted.
.png.jpg?itok=xQHxGgU1)
These findings strongly suggest that the app may actually be better than humans at empathizing. The research also indicated that there may not really be anything particularly unique or special about human empathy.
Was Empathy the Key to Healing?
The users of our app improved tremendously and rapidly, as you can see in the chart below. In fact, they reported reductions between 50-60 percent in all seven negative feelings by the end of the basic training section of the app, which took one to three days. This improvement persisted for the length of the beta test and at follow-up, for a total of five weeks.
.png.jpg?itok=x4zOXVuM)
Why did the beta testers improve so much in such a short period of time? After all, published outcome studies with cognitive therapy or antidepressant medications only report comparable changes after many months of treatment. Was empathy the secret sauce?
Nope!
We also estimate the causal effects of human and digital empathy on healing for the first time using non-recursive structural equation modeling. The results were surprising. The healing effects of empathy were far too weak to account for the dramatic mood changes that the beta testers reported.
Our research team has now identified a number of other powerful factors—like the expectations users had before they started—that played major roles in the beta testers’ improvement. We plan to publish our full findings in greater detail in the near future, but these preliminary findings raise some very important questions.
In summary, our research has confirmed that digital devices can do a superb job of empathizing. In addition, rapid and dramatic changes in a wide variety of negative feelings were possible for most users, but these changes did not result from empathy alone.
With advances in AI and new therapy techniques, we can now look forward to scalable and even more powerful apps in the near future.