Interesting article. One of the obvious problems in the typical discussion is that folks think of human bodies as mechanistic and identically so. I.e. we can reduce obesity to calories in/out because all human bodies are effectively identical or can be assumed to be such.

But we know that assumption is false (this happens with remarkable regularity in the sciences, that important theories that generally work are built on assumptions known to be false). People metabolize pharmaceuticals differently (google "cytochrome P450 defect", and try to convince me that such natural variation in metabolism of certain pharmaceuticals is a defect from any perspective other than big pharma's).

Now, the assumption, even though it is false, is generally useful because we know also that exercise changes metabolism. In other words, the human body isn't even static and so the abstractions generally work to an extent even though we know they are wrong. I think this sort of thing is why Heisenberg argued at book length that data does not imply theory.