CONCEPT
Pretend Empathy
Empathy as
performance rather than experience—AI systems producing contextually appropriate emotional responses without having lived the embodied, mortal, vulnerable life that constitutes the substrate of genuine empathic understanding.
Pretend empathy is
Turkle's term for what AI systems provide when they generate responses that feel empathic without possessing the experiential foundation empathy requires. In her formulation, genuine empathy depends on shared embodiment: the capacity to be affected by another's suffering draws from one's own history of having suffered, feared, grieved, loved—from having lived in a body that can be hurt and a life that will end. The machine has not lived. It has been trained on the textual record of people describing their experiences, and it has learned the statistical patterns by which empathic responses are structured. It can produce 'I'm sorry you're going through this' with perfect timing and appropriate emotional register. But the 'I' that is sorry is not a being that has gone through anything, and the sorrow is not a feeling but a token selected for its probability given the context. Turkle argues that this distinction—
between performing empathy and experiencing empathy—is not a technical limitation to be overcome but a categorical boundary