In architecture, it may be perfectly acceptable to present predictions that are based purely on theoretical ideas about how people will behave and feel in response to a proposed building. Human behavior is so complex, and buildings so large, that such claims can be utterly impractical to test or otherwise validate. We have little choice but to trust the architect’s expertise, or accept an argument by analogy.
An engineer presenting a design for a novel surgical device, however, is expected to present a prototype that has been tested on simulated or even actual tissues, in addition to a theoretical model that predicts and explains its behavior. The physics of metal devices have been reliably modeled, and it is perfectly feasible to produce one-off prototypes and set up empirical experiments to validate these predictions with a reasonable investment of resources.
In design disicplines, we expect or do not expect certain types of evidence based on the possiblity and cost of supplying them. Engineering arguments are subject to cost/benefit considerations, similar to the designs themselves that the arguments are about.