-
+100 Since my own PhD days, I have seen this as one of the worst parts of being in computer science academia (there are many good ones, but this one is deeply frustrating). Now, as a member of (too many) PCs, I keep losing arguments over accepting papers because of "novelty". 1/ @vardi/1562932318708310016On twitter.com
❤️ 6 Favorites Mood -4 🙁
-
Now, the point of "novelty", as @vardi points out so well, is that it is one of the most subjective aspects to judge, and yet seems to be an objective criterion. It should be easy to evaluate, right? 2/Permalink On twitter.com
-
@vardi If a sufficiently similar method, experiment, system, etc. has been published before, it's no longer novel. But let's dive into that: 1) define "sufficiently similar" 2) you actually have to know the previous work 3) all other reviewers should also read the previous work. 3/Permalink On twitter.com
-
And there's simply soooo much of previous work. No "domain expert" will or can know it all. And no, the other reviewers will not (be able to) take the time to read all the supposedly similar work another one mentions in their review (which might well be their own paper...). 4/Permalink On twitter.com
-
So the other reviewers will take the word of the first one stating that there is previous work and it's therefore not sufficiently novel. Case closed, one paper off the pile to discuss. And it's close to impossible for authors to address in a rebuttal. 5/Permalink On twitter.com
-
Summarizing, the subjective and often very poor judgement of "novelty" as criterion for acceptance of computer science papers has been holding the whole discipline back for decades. Let's address this one. (And then we start talking about "soundness".) 6/6Permalink On twitter.com