Saethydd thanks for an actual response. You've largely opted for the "they are not real knowledge" option. Apart from sociology, which describe as a science. I think you could only really maintain that sociology is a science by stripping out many qualitative approaches that are commonly viewed as being key parts of sociology: discourse analysis, semiotics, ethnography, action research and so on. Sociology isn't a branch of psychology. "Margin of error" in sociology is a term that's only applicable to some quantitative approaches of the discipline.
The problem with the proposition that science offers "knowledge" whereas other disciplines only deliver "beliefs" is twofold. On the one hand it ignores the fact that science is empirical and is therefore based on inductive reasoning. And inductive reasoning cannot deliver definitive "knowledge" since its propositions are always provisional and by definition vulnerable to future disconfirmation. The other problem is the idea that different subjective approaches to knowledge don't count as knowledge presents an impoverished and unworkable view of the world.
Newtonian physics may be wrong in some particulars for example, but in many practical situations we'd be glad to use it rather than guessing.
Similarly history may not always deliver reliable results, but in the face of Holocaust deniers we'd be glad to rely on its approach to knowledge than abandon any sort of knowledge about the past.
Science and non-science approaches to knowledge turn out not to be all that different after all. They all rely on unprovable assumptions (in the case of science, induction), they are all subject to theoretical developments, and they are all only as good as the practical uses to which they can be put.