There is a proliferation of predictive models and their outcomes vary widely. LA Times consistently predicted a Trump win, FiveThirtyEight was down to a coin toss a few times, while The New York Times was persistently pro-Hillary. While those models were meant to help people make sense of the large number of individual polls, now we need models to make sense of the models.
It is better than what we do, informally. We all use models (of some sort) when we process information; e.g. I discount certain publications and value others. But few people actually make their models explicit and ask others to critique them.
Making a model public and explicit is a great first step. Building more models on top is a good thing, not a bad thing.
