To do this, the researchers used a
simple trick called the rubber hand illusion: A person places her hand on a table where it is
hidden from
view in a box or under a smock.
Because Schwartz's model is
simpler it is easier to account for and quantify the uncertainty in it (in fact much of the uncertainty in complex GCMs is
hidden eg see Stainford et al referenced in the post), so if you take the
view that you are interested not just in the mean but the variation in the estimate Schwartz's model, despite being
simpler, gives you better information.