Coming from the natural sciences, I typically see a scientific model as follows: A scientific model takes all available evidence and attempts to ascribe a relating structure to all this information that will allow one to make a set of predictions about the natural world. When new evidence is introduced or when the model fails to accurately predict universally, a new model must be developed or an existing one must be modified. Because of the inherent complexity of the world, a model has to compress information through structure. It cannot simply describe everything - it has to have sufficient structure such that very complex processes can be explained through simpler systems. This is a very general definition and can vary if you are talking about mathematical/statistical models or biological models.
In less quantitative fields such as philosophy/social sciences, it seems to me things are different. Many philosophers seem to start with observations or subjective "axioms" (sometimes including empirical data) about the world and then attempt to find an underlying structure that relates these observations. The resulting model is not always testable. It may simply collect observations together to try to unify them. Many philosophers that have done this an invent very detailed terminology to accomplish this (many of the Continental Philosophers such as Hegel, Sartre, and Heidegger come to mind). In fact, even in this post I am attempting to define how other philosopher develop models through my own personal observations of how I have seen models developed in other authors I have read.
In many cases, I will read about an author's personal model and disagree entirely about how the observations are organized. This can happen in the case of scientific models too (after all, the results may be objective but the interpretation can be subjective) but because it is working on testable hypotheses, it is possible to evaluate the model's success or failure to predict reality.
My question: Are there any standards for how models are developed in philosophy for rigor and accuracy? What guidelines are typically followed when developing a model? How does one ensure that the model is accurate universally and not subjective when it cannot be tested in a formal, scientific sense? How has this been treated historically (1850-1900 science) versus the modern scientific era?