Creating knowledge is learning. It is done by guesses and criticism, also called conjectures and refutations. The guesses can be anything at all. There's no restrictions or limits on them. But the criticism narrows down what guesses will work well. It's important to narrow things down because most possible ideas aren't true, and we're looking for the truth.
If a guess is criticized, then either it's mistaken, the criticism is mistaken, or they both are mistaken. Many guesses will be mistaken and we shouldn't mind that. Whenever we refute a guess, we've learned something. When we refute a criticism, we've also learned something. Ideas are always failing, but in the process we learn. Over time we find more of our mistakes and move on to better ideas.
Some theories about how knowledge is created try to create ideas in such a way that they will turn out good. They focus on the method of creating ideas. But actually, how ideas are created isn't important. The criticism does the heavy lifting. And if any bad guesses are created, it's no problem at all, they will soon be refuted by criticism. (If no one can think of a criticism of it, then it's not a bad guess, it's a pretty good guess worth considering.)
As we build up more and more criticism in a field, it gets harder to make new guesses that aren't already refuted by existing criticisms. Knowing criticisms to apply to any new ideas puts restrictions on what guesses will be able to survive initial consideration. In this way, guesses become progressively less frequent or less ambitious. We narrow in on the truth and change more slowly.
Criticism can be thought of as a filter. Each criticism blocks out guesses that do or don't meet certain criteria. With enough filters in place, any idea that makes it through for further consideration may already be pretty good. That's why the guesses that get to people's conscious mind are far superior to random guesses — they are going through numerous filters in the unconscious mind which limit what they can be.
What's good about filters is that they help guide us through infinite possibilities. We couldn't handle the task without them. The downside of filters is they can create blind spots. For that reason it's important to be aware of what filters or criticism one is using, and to be able to update them if they cause problems. We should aim for thoughtful criticism not stubborn criticism.
Building up good, generic criticisms is itself a type of knowledge creation. It's learning about what categories of guesses are mistaken, and it helps us avoid making those mistakes in the future. This is only possible if we keep an open mind about what types of criticism are useful, and use thinking processes that allow for change.
In science, evidence plays a major role. Evidence is used to rule out any theories that it contradicts. We form criticisms like, "Theory X predicts Y, but we did an experiment and got Z not Y. Therefore X is mistaken." In this way we use evidence in our scientific criticisms.