The Plural Self — Summary
You are not one person. You are Margaret-the-grandmother when your daughter visits with the children, Margaret-the-patient when the physician enters, Margaret-the-widow when alone at night. Each is real. None is more authentic than the others. The selves are not masks over a single true self — they are each genuinely you, constituted by the relationships that call them forth.
Most AI systems model humans as unified agents with stable preferences. This works adequately for simple domains but systematically fails for anything that matters. Ask Margaret what she wants for dinner: the answer depends entirely on which Margaret you are asking. These are not different answers to the same question — they are different questions asked of different selves. A system averaging across contexts produces recommendations suited to no actual context.
Understanding the plural self would require identity detection (recognizing which self is active), distinct modeling (not one model with contextual adjustments but genuinely separate models), transition awareness (tracking when identity shifts mid-interaction), and most challengingly, relational constitution — grasping that the grandmother-self does not pre-exist in storage waiting to be retrieved, but comes into being in living relationship with the grandchildren. Context does not activate identity; it constitutes it.
An AI system could learn to route between Margaret-configurations without understanding any of this. Pattern matching on contextual cues, separate preference databases, transition state management — all of this could work functionally while the system understands nothing about what it means to be a grandmother. The functional-phenomenal gap runs through identity itself.
The pragmatic response is that functional modeling is enough if it enables better assistance. But the pragmatic response assumes we know what “helping” means without understanding. The grandmother does not need meal recommendations optimized for nutrition. She needs to feed her grandchildren in a way that expresses her love and creates memories. A system that cannot grasp grandmotherhood will optimize for the wrong thing. Understanding is not a philosophical luxury. It is what makes the difference between help and its simulation.