Empathy is increasingly held up as the one leadership quality AI cannot replicate. The irony, then, is that new research from Stanford suggests it can be trained using AI – but that doing so may produce leaders who communicate empathy without actually feeling it.
Anna Queiroz, now an associate professor at the University of Miami, conducted a studyin which participants used virtual reality to practice a performance review conversation with a virtual employee (themselves). After six minutes in a headset, something measurable changed. Participants started using more first-person language—“I understand what you’re telling me” rather than “you should do this”—and introduced more emotional words into their speech. By the markers researchers use to assess empathetic communication, they had improved.
There was just one problem. When asked how empathetic they actually felt toward the virtual employee, the answer was unchanged. The behavior had shifted. The feeling had not.
What The Research Actually Shows
Queiroz’s study placed participants across three conditions. In the first, they played a manager giving feedback to a virtual employee—establishing a baseline. In the second, they then embodied the employee and watched themselves as manager. In the third, they did the same but also received structured feedback on whether their communication demonstrated the five elements of Queiroz’s LIVES framework: listening, being incisive, validating the other person’s experience, expressing their own emotions, and offering support.
Using natural language processing to analyze the transcripts, Queiroz found that participants who received feedback showed a meaningful shift in their second interaction. Rather than directing, they were including themselves in the conversation. The communication had become measurably more empathetic.
And yet, participants’ self-reported empathetic concern did not change across conditions. As Queiroz told me, “we can change our communication even though we don’t feel more empathy towards the person. So it can be practiced.”
But practiced to what end? Queiroz explains that VR can be used in two ways: to build genuine empathy by embodying another person’s lived experience, or to train how to communicate empathetically. Her study did the latter. The question is: if the communication is indistinguishable from the real thing, does the feeling behind it actually matter?
Why This Matters Now
Siduri Poli, partner and CMO at 0TO9, a European fintech venture builder and investor, would not back a founder she doesn’t believe has genuine empathy. “AI outperforms human IQ, but not EQ [emotional intelligence],” she told me. “Humans are still dominating in EQ.”
In her work supporting early-stage founders, Poli describes moments where what she says could determine whether a company survives. A founder who is on the edge of giving up needs to be “truly seen.” She is direct about what that requires: “That is something that only a human can make you feel.”
Therefore, empathy is not a soft skill being added to leadership programs as a nice-to-have. It is increasingly the capability that AI cannot replicate—and is a key soft skill that, as I have discussed in previous articles on this topic, matters enormously when creating high performing teams. Poli is blunt about whether empathy can be faked: “I think it’s hard to fake empathy. But the question is not how to fake it. The question is how to become really good at it.”
The Opportunity—And The Risk
For most of the modern workplace’s history, consistency, efficiency and predictability were the qualities organizations rewarded. Being human, in the messy, relational sense, was something to be managed rather than expressed. AI is more consistent, efficient and predictable than any person will ever be.
Poli frames this not as a threat but as a genuine release. “Back in the days when we didn’t have AI, maybe people needed to step up and become the robot. But now we actually have robots being robots, and that leaves room for us to actually be people again.” She sees it as an opportunity to return to the connection and humanity that makes leadership and work meaningful. Though, worryingly, many will not have that option.
However, Queiroz’s findings raise the question as to whether genuine connection is necessary. If VR can shift how people communicate empathy without shifting how they feel it, then the same could apply at a much broader level. Organizations could respond to the AI era by coaching leaders in the language and behaviors of empathy while the actual capacity for human connection remains undeveloped.
As I have stated above, empathy will become one of the key skills within an AI dominated workplace. The question is whether organizations will invest in developing the real thing or settle for a performance?
