Why won't good journals retract papers that were proven wrong?

Question

I saw some research papers still present in good and reputed journals even though they were proven wrong by other papers. The subject is cryptography.

If a paper is proven wrong, it does not contribute a valid result to the corresponding topic. What is the reason for keeping them intact, which can cause issues for newbies?

Answer 1

Without knowing the specifics of the papers you are referring to, it's difficult to respond exactly. In general, however, there is more than one way for a paper to be "wrong", and most "wrong" papers should not be retracted.

The thing is, a paper typically contains more than just a single assertion. When you say that a paper is "proven wrong", it sounds like you are referring to the high-level conclusion, e.g., "Cryptographic protocol X is secure against replay attacks." Then somebody else comes along and shows that actually, there is a way to do a replay attack after all.

Such a refutation, however, often does not actually invalidate any of the actual technical methods or results presented in the original paper. Instead, it will more typically involve showing that they were insufficient in some way for supporting the high-level conclusion. For example, the problem statement might have been formulated too narrowly, or the authors might have drawn a conclusion that was stronger than their evidence actually supported.

In such a case the conclusion of the paper may indeed be wrong, but everything else is indeed correct. Careful phrasing by the original author may in fact mean that the conclusion is even still technically correct (e.g., "According to this formulation, cryptographic protocol X is secure against replay attacks."). In short, the results stand but their implications are much less than the original investigator may have believed.

Bottom line: many refutations may be understood as changing our interpretation of results derived using valid methods, while retraction is generally reserved for invalidating flaws in the methods themselves.

Answer 2

Think of the history of scientific papers as being something like an informally maintained blockchain. The integrity of all depends on the continued presence of the old things whether they have been superseded or not. Even if they have been superseded for errors. The reason for this is that things get referenced in newer work. The new things may be corrections or may, in fact, rest on a shaky ground. But future researchers need to be able to resurrect all of the contextual history of ideas and their expression so that science can advance without needing to start over whenever something is found wanting.

In fact, if an old paper is made to somehow disappear and a newer one covers the same ground, but correctly, then the newer authors are, arguably subject to claims of plagiarism since they "used" some of the old words and ideas that they weren't able to find by searching the literature. There are likely old versions of the (incorrect) paper cached in various places. Better for the old paper to remain in place so that references to it, especially correcting references, remain valid.

In some fields, having access to old, incorrect, papers can be especially valuable to a student. If the superstars of a field go wrong in proving something important, it is useful to know how they went wrong. That way, similar errors can possibly be avoided in the future. In analysis, for example proofs that contain lots of conditional clauses with quantifiers (for every, there exists, ...) can easily get out of control and hard to follow. Many of these have actually occurred in the literature. It can be fun to find them, and not so easy, in many cases, to correct them.

Answer 3

In many fields, and probably especially cryptography, the evidence of an incorrect result can be very relevant either:

1) to reduce the effort in a “wrong” or incorrect or less-fruitfull direction,

2) to provide an impetus towards a solution or a solution for another direction or method.