I’ll bite the bullet: if a robot has a complicated enough internal representation of the world, it may very well develop a concept of love (or “care”, or “noticing”, or “intention”. Love is such a slippery word…) that we would have to trust.

Imagine a cat-sitting robot. The robot can differentiate between individual cats. It learns how to play with the cats and feed them in in their preferred way. The cats grow to trust the robot and enjoy its company. When the cats become sick and old the robot knows how to help them and ease their pain. Over decades The robot remembers cats in its care that have died, and new cats spark recognition of previous cats it has known. It becomes better at caring for a wider range of cats as its experience grows. The cats cry out when it leaves. When there are no cats around the robot remains motionless, but springs into action and play as soon as cats are around. Children would describe the robot as “happy”.

If after some decades I smash it with a hammer and recycle the pieces, am I killing something? Are its internal representations and control systems not a kind of thing that produces “qualia”?

What's really a head trip is that I don't actually know another human is experiencing grief either. They could be a sociopath and not actually feel emotions, but are pretending to in order to benefit them in some way.