[guntars]

In the beginning the author of the comment said that a broken model results in a loss of confidence which then prevents us from acting on that model and forces us to figure out a better one. They're saying that's what depression is - the downtime when you're supposed to play it safe and re-analyze the data and build a better fitting model. I can see how it can be better to not act on a broken model in case it could result in some harm, but it also stops us from gathering new data that might be important.