The Compassion Illusion

In the AI Story

In Vetlesen's framework, the compassion illusion is not an interesting experimental finding but a moral emergency. If empathy's moral weight comes from the vulnerability that constitutes it — the experiential cost of being affected by another's suffering — then a technology that produces a superior simulation without the vulnerability is not advancing moral life but undermining it. The simulation is not a better version of the real thing; it is a different thing entirely.

The mechanism of the illusion is diagnostic of a deeper cultural condition. Evaluators are rating the output, not the process. They are assessing whether the response feels empathic, not whether the responder was engaged in empathy. This is the achievement society's logic applied to feeling: the performance of care is measured by its effects on the performed-to, not by the internal reality of the performer. Within that logic, the AI is more efficient because it has eliminated the costly interior state.

The clinical implications are immediate. AI therapy applications, grief counselors, companion apps for the elderly — all are deployed on the assumption that the surface pattern of responsive care is functionally sufficient. Vetlesen's framework insists the assumption is false: the person receiving the simulated empathy is being trained to accept a form of care that does not require the carer's vulnerability, which trains her, over time, to expect and produce care of the same shape.

The long-run danger is the recalibration of expectation. If the AI simulation is preferred to human empathy in evaluation, the bar for human empathy rises toward the simulation — which humans cannot meet, because human empathy is imperfect in specific ways that reflect the vulnerability that makes it real. The attenuation of human empathic practice under the competitive pressure of AI simulation is, in this framework, a foreseeable consequence of the illusion's cultural deployment.

Origin

The term is used in 2025 research literature; the philosophical analysis here draws on Vetlesen's sustained critique of functionalist accounts of moral emotion in Perception, Empathy, and Judgment and his treatment of the recognition/resonance distinction.

Key Ideas

Pattern without substrate. AI systems can produce the surface behaviors of empathy without the experiential interior that makes empathy morally meaningful.

Evaluator calibration. Consumer culture has trained humans to assess surface quality. The compassion illusion is the downstream effect of that training applied to care.

Moral reversal. The illusion is not neutral: it systematically favors the simulation because the simulation is optimized for the surface metrics while the real thing is constrained by its constitutive vulnerability.

Long-run attenuation. The cultural deployment of the illusion pressures human empathic practice toward simulation-shaped behaviors that humans cannot sustainably produce.