But how does a human arrive at the belief that some formal system, e.g. PA, is “really” talking about the integers, and some other system, e.g. PA+¬Con(PA), isn’t? Can we formalize the reasoning that leads us to such conclusions?
Well, human beings are abductive and computational reasoners anyway. I think our mental representation of the natural numbers is much closer to being the domain-theoretic definition of the natural numbers as the least fixed point of a finite set of constructors.
Note how least fixed-points and abductive, computational reasoning fit together: a sensible complexity prior over computational models is going to assign the most probability mass to the simplest model, which is going to be the least-fixed-point model, which is the standard model (because nonstandard naturals will require additional bits of information to specify their constructors).
A similar procedure, but with a coinductive (greatest-fixed-point) definition that involves naturals as parameters to the data constructors, will give you the real numbers.
Well, human beings are abductive and computational reasoners anyway. I think our mental representation of the natural numbers is much closer to being the domain-theoretic definition of the natural numbers as the least fixed point of a finite set of constructors.
Note how least fixed-points and abductive, computational reasoning fit together: a sensible complexity prior over computational models is going to assign the most probability mass to the simplest model, which is going to be the least-fixed-point model, which is the standard model (because nonstandard naturals will require additional bits of information to specify their constructors).
A similar procedure, but with a coinductive (greatest-fixed-point) definition that involves naturals as parameters to the data constructors, will give you the real numbers.