Why Machines Cannot Truly Feel: The Empathy Gap in AI

Artificial intelligence systems can be programmed to say the right words when a user is upset, but a new review argues that what they are doing falls far short of genuine empathy.

The authors begin from a well-established premise: empathy is linked to a range of positive outcomes, including health, morality, positive interpersonal relationships, group cooperation and social harmony. But empathy, they stress, is not a single skill. It is a complex socioemotional experience that unfolds through dynamic, interactive cognitive, affective, motivational and behavioural processes. That multidimensional structure is precisely what makes it difficult to reproduce in a machine.

The positive potential of AI agents as human companions is, in the authors’ view, laudable to pursue, yet it carries serious risks and challenges. Despite those risks, engineers have committed themselves to developing empathy capacity in AI agents. Scholars have already noted that such agents can simulate some aspects of empathy, most notably cognitive empathy, the ability to represent or model another person’s perspective. What they cannot do, the review argues, is simulate authentic empathy, a limitation that has become known as the empathy gap.

The essay extends this discussion in three directions. First, the authors draw attention to the biological, psychological, interpersonal and developmental processes that pose challenges for any attempt to make AI feel authentic empathy for a human. Empathy, on this account, is grounded in embodied and developmental histories that a simulated system does not share, so reproducing its surface expressions does not reproduce its underlying experience.

Second, the authors assert that any serious effort to simulate authentic AI empathy for a human must account for culture-group processes and differences.

Third, the authors present a heuristic model of empathy intended to highlight the complexities of simulating authentic AI empathy for humans.

The review closes with implications about the dangers and limits of AI human empathy, with a particular warning: generative AI interactions with children are proliferating. If young users encounter systems that appear empathic but lack authentic empathic capacity, the gap between simulated and genuine care becomes a developmental and ethical concern rather than a purely technical one.

Subject of Research: The gaps and risks in simulating authentic empathy in artificial intelligence

Article Title: Complexities, gaps, and risks in simulating AI empathy

Article References: Carlo, G., Ahn, J., & Cervera, K. (2026). Complexities, gaps, and risks in simulating AI empathy. AI, Brain and Child, 2(1), Article 13. https://doi.org/10.1007/s44436-026-00037-3

Image Credits: AI Generated

DOI: 10.1007/s44436-026-00037-3

Keywords: AI empathy, empathy gap, cognitive empathy, AI companions, generative AI, child development, socioemotional processes, culture, human-AI interaction, AI ethics, review essay, AI, Brain and Child