Depends. If I believe the function should be linear, then I use the first idea. But, when the underlying physics are non-linear, I might discuss why I think so, and what form the non-linear equation might have, if it's germane to do so (i.e. in discussing the limits of a linearized model).