When a journal is looking for "originality/novelty/newness," what are they looking for? Sure, this is a somewhat subjective question and open to personal opinion, but I think there are indeed cultural standards in the field which give some level of broad consensus, say, among editors at top journals.
What I can think of that would satisfy such adjectives would be things like:
- Solving an unsolved problem in an unexpected way.
- Creating a new concept that is useful for some problem.
- Solving an problem for which a solution already exists in a new and unexpected way.
But there has to be something else to it, e.g. the new method or concept used must meet some technical proficiency bar or something.
To what degree is extending previous results "novel?" Most research in every subject is extending previous results. Very rarely does something completely and exceptionally new and unexpected come around. Therefore, there must be some reasonable (even if fuzzy) bar set somewhere in the middle.
I'll give my example. I extended a (stochastic process/probability theory) result to a very large class of models using a method the original author did not use. This resulted on improved accuracy of nearly every result in the original paper and easier proofs. In some cases I get exact formulae where the original only had approximations (for that one model considered there, but mine now applies to an unlimited class of models).
I claim:
- using methods that the previous paper did not use (nor any papers citing that original result)---I think it is two subfields that don't communicate much (but are related in obvious ways);
- improving nearly every result;
- for a much larger class of models;
- inventing at least one new concept which was necessary for the result; should at least be somewhat close to the bar for novelty for a top journal. I know I'm aiming high and failure is the most likely end result!
Part of the problem is that my proofs are very easy (due to the use of said methods). The only hard parts using high powered machinery are already well-established. The most novel part is really in developing the new concept that allows the result and placing using methods of this other subfield.
So, yes, part of this is me griping, but the real question is the first one stated at the outset.
What to top mathematics journals think of as new/novel/original? I mean the hardest journals to get into in any arbitrary field, by whatever your favorite metric is.