Do children "play" with dolls? I don't think that is a slam-dunk. As far as we know, children may be mindless automata and what we call "doll-play" may be unfeeling mirroring of adult anthropomorphizing behavior absorbed through something like blindsight.
> You're kind of making my point for me here.
Not at all. My point was that calling a puddle "melted ice" has nothing to do with mistaking it for any kind of ice or ascribing characteristics of ice to it. Likewise with "dead people".
> A baby's response to pain is all but indistinguishable from its being, say, hungry, or just in a bad mood
That's simply untrue, even in neonates (not to mention babies several months older). Look into the abundant research applying the Facial Action Coding System and derived systems to infants and their pain vs. other distress responses.
> You can't conclude anything about subjective experience from I/O behavior alone.
This is a denkverbot, not productive skepticism. We can draw robust, valuable conclusions about subjective states without enjoying deductive certainty.