Summary: A 2026 meta-analysis of 24 trials found that digital empathy interventions show a small initial effect, but it vanishes after correcting for publication bias and fails to beat active controls. The evidence raises serious doubts about whether screens can genuinely teach us to care.
Ten years ago, the idea that an app or online program could make you more empathetic sounded like science fiction. Today, digital empathy interventions are everywhere, from workplace training modules to therapeutic platforms. But do they actually work, or do they just make us feel like we tried?
The Digital Empathy Evidence Gap
A new meta-analysis published in February 2026 in Current Psychology offers the most rigorous look at this question so far. Researchers pooled data from 24 randomized controlled trials involving 3,137 adults, producing 26 comparisons of digital empathy interventions against various control conditions.
On the surface, the results look cautiously optimistic. The overall effect size came in at g = 0.19, which is statistically significant (95% CI [0.05, 0.32], p = .006). That translates to a small positive effect. Small, but real, at least at first glance.
But here is where it gets uncomfortable.
Why the Numbers Fall Apart
When researchers applied a trim-and-fill correction to account for publication bias, that small effect collapsed. The adjusted effect size dropped to g = 0.07 (95% CI [-0.07, 0.22], p = .322). That is no longer statistically significant. In plain language, once you factor in the strong likelihood that negative results went unpublished, the overall evidence says these interventions probably do not work.
The finding gets even more damning when you look at how the studies were designed. Digital interventions did not outperform active control conditions. An active control is basically a placebo activity that feels similar to the real intervention but lacks the key ingredient. If a digital empathy tool cannot beat a placebo, it is hard to argue it is doing anything meaningful.
The One Glimmer That Holds Up
There is a partial exception worth noting. Studies that compared digital interventions against non-specific control groups, like waitlists or no treatment at all, did show a small positive effect of g = 0.22 (95% CI [0.12, 0.33], p < .001), and this result held up after trim-and-fill correction.
So digital interventions appear better than doing nothing. But that is a low bar. Anything that engages you with the topic of empathy for a while will likely outperform sitting on a waitlist.
Follow-up data from six studies told a slightly more encouraging story, with an effect size of g = 0.29 (95% CI [0.07, 0.5], p = .008) that also survived bias correction. This hints that some effects might grow over time, but with only six studies providing follow-up data, that conclusion is fragile.
Notably, the meta-analysis found no significant moderators. That means researchers could not identify any factor, like intervention type or format, that reliably predicted better outcomes.
What We Still Cannot See
This analysis leaves enormous blind spots. The available abstract does not detail what kinds of digital interventions were included, whether VR scenarios, mobile apps, or video-based training. We also have no data on whether cognitive empathy (understanding someone's perspective) and affective empathy (feeling what someone feels) respond differently to digital formats.
The meta-analysis authors themselves point to a direction forward: more advanced technology that can offer attuned interaction mimicking a human relationship, along with real-time feedback for participants. That kind of capability barely exists today, which tells you how early we still are.
The Uncomfortable Takeaway
The most honest reading of the evidence is this: digital empathy interventions have not yet proven they can do what their creators claim. They look better than nothing, but not better than a well-designed placebo. And once you correct for the file-drawer problem where unsuccessful studies never see print, even that thin advantage disappears.
This does not mean technology is hopeless for building empathy. It means the current generation of tools is not there yet, and the published literature has been painting an overly rosy picture.
So the next time your employer assigns a digital empathy module, ask yourself a simple question: would you trust a medication that only works because the failed trials were never published?
Comments