1

This is not necessarily a TeX meta question but I do not know where to ask it. From the arXiv I am used that submissions get flagged if they have substantial or some overlap with papers by the same or other authors, see here. There the algorithm is rather simple, and extremely helpful because every once in a while one reads

Substantial text overlap with paper by other authors.

The interpretation is then (rather) obvious. (Yes, there have been problems where author A wrote a report, B copied it but was the first one to post it on the arXiv, A recycled their own paper and got into major problems because it now said that their paper had substantial overlap with the submission of B. Let us not talk about such subtleties here.) My question is whether it is feasible to have something similar for our main site. Ideally it would not only check text overlap but also code overlap. Are there any ideas to implement such features on this site?

Clarification (since at least one user does not seem to have gotten the point): this is (obviously) a question on the overlap of answers. For questions there is a mechanism in place, one can mark the copied question as duplicate. In fact, when typing a question, there is already a robot that identifies related questions on the spot. And such lists appear for all questions. So this site does have software that allows one to identify potentially related questions. (Whether or not the suggestions made by this software is always optimal is something I cannot tell.)

My question in more technical terms:

Can one make the software that identifies related questions also list related answers, and attach these lists to the respective answers?

P.S. I understand that this is not fundamental research, and in a way you can find most of the things somewhere in some manual. Yet it would be nice if there would be some means to make users indicate the source of their codes and strategies. That way there will be crosslinks to other posts, which will add value to the answers because if everyone would add these links, the related posts are just a few mouse clicks away.

  • 1
    How should something like this be implemented? For example, once the answerer clicks "post", a pop-up shows up that states something like "Substantial text overlap with paper by other answerers." with possible related answers. Can the answerer just ignore that? Or how would they be able to move forward? – Werner Jul 12 at 21:31
  • 1
    @Werner Please read my question: "Can one make the software that identifies related questions also list related answers, and attach these lists to the respective answers?" – user121799 Jul 13 at 4:44
  • 1
    @Downvoters could you care to explain why this question is down-voted? – Raaja Jul 13 at 5:28
  • 1
    @Raaja Sorry, you're right, maybe the reason for my downvote is not enough clear from my answer. Legally or ethically the topic is very important, but for the aim of the site (provide good solutions) it is not so important as to put a complicated mechanism to avoid it. – CarLaTeX Jul 13 at 10:42
  • 1
    @CarLaTeX thanks for the clarification. I have not much knowledge on the licenses so as to speak. Based on this tex.meta.stackexchange.com/questions/1255/…, how I understand these complicated terminology is this: it doesn't matter whether we have our yard wide-pen/with barbed-wire/fenced/gated. But it's obligatory to explicitly specify where the code is adopted from (unless the user explicitly says no). A piece of script is one such a thing what a developer would be proud of imo. – Raaja Jul 13 at 12:26
  • 1
    @Raaja Yes, the standard license of TeX.SE is "you can use the code but you must say where you copied it". – CarLaTeX Jul 13 at 12:30
  • 1
    @Raaja The main intention of this request is not academic honesty, even though this is important to me. To give you an example: I was writing an answer on cuboids (which were referred to the OP as cubes). Wouldn't it be great if anyone looking at this answer would also see a survey of other answers on this topic? As of now, users who are still active can add a comment: "related: blabla", which may be viewed as obnoxious, or, in case of no longer active users like Jake their answer, which may well be superior, will just be ignored. – user121799 Jul 13 at 12:32
  • 1
    [Continued] ofcourse, in the spirit of this site, we can say it's ok as it is mainly to provide good solutions. But it doesn't anyway should promote unethical behaviour. From what I see, sometimes good questions help us generate new packages from very great answers. Given such circumstances, if someone else takes the credit the original developer would be devastated (so as to speak). And that, is not an example of fairness. After all we are here for learning/having fun/helping others. So in conclusion,I think academic honesty is very much important for the fairness of this site :-). – Raaja Jul 13 at 12:32
  • 1
    @Raaja I also agree with that, and disagree with the assertion that disclosure of the source of the code is not relevant for the quality of the posts. In addition, seeing related answers will help both the answerer and others to improve the posts. – user121799 Jul 13 at 12:34
  • 1
    @marmot thanks for clarification, I see this part. However how I see the impact is that such a feature will be far more useful in solving the long-going topic on academic-honesty as well. – Raaja Jul 13 at 12:36
  • 1
    @Raaja Academic honesty is hard to achieve in full glory because most of the stuff we do here has done in some variation before, and it may not be possible to give full credit to everyone who deserves it. But yes, we do have a huge room for improvement. As of now, it seems that you get penalized if you link related posts (perhaps because users then feel: oh, that answer is just a variation), but if you just copy or claim you have fixed some code or make other claims, this will be rewarded. (Did I say I disagree? If so, this was not intended. I disagreed with an assertion that you did not make) – user121799 Jul 13 at 12:42
  • 1
    @CarLaTeX "you can use the code but you must say where you copied it" already exceeds what I am asking for. – user121799 Jul 13 at 12:43
  • 1
    @marmot I think that is the rule (legally) – CarLaTeX Jul 13 at 12:47
  • 1
    @CarLaTeX Well, it is certainly not an enforced rule, is it? If it was, then some users would be in serious trouble. – user121799 Jul 13 at 12:49
  • 1
    @CarLaTeX Yes, but this is not about trials. We have questions on whether it is OK to unaccept an answer, so why can't we have no discussion on what's wrong with copying codes without disclosing the source? – user121799 Jul 13 at 12:54
9

Sorry if I answer with a question, but why is academic honesty so important to you?

Please don't misunderstand me, but if someone copies my code means that my code is good. Of course, I would be happier if they cited me (it is also a rule of the site, I think) but the most important thing is that users get the answers they want.

An adage amog old Italian COBOL programmers says: "The only one who wrote a COBOL program is Spoletini, anyone else did only cut and paste" (prof. Spoletini wrote a famous COBOL manual).

I more or less copied from manuals, guides, TeX.SE answers, every line of code present in my documents. Honestly, after years, I would be very difficult to me to give the correct attribution to them.

It is surely very unfair is someone copies an answer tel quel, but I think it's enough to put a comment that says "Your answer seems to be copied from ..., it would be fair if you give the correct attribution". For the most serious cases there is also downvoting and/or flagging for moderators' attention.

  • There are many reasons why I think this is important, a practical one is mentioned in the P.S. of my question. If the posts that copy from each other are linked, it will be easier for the user to find related answers. Another reason is that if everything that is somewhat original gets copied without attribution, users may be less willing to contribute original ideas. Last but not least this is just fair play IMHO. And the reason for my question is that I do not see why we should dump everything on the moderators. – user121799 Jul 12 at 4:59
  • 1
    @marmot You can simply comment and downvote. Flag for mod's attention only in serious cases. If you put the link in the comment it appers as related question on the right side. – CarLaTeX Jul 12 at 5:05
  • Well, this always creates tension and a bad taste. If user A posts a comment saying that user B copied from C\ne A, then B in some cases gets mad, reacts with revenge downvotes. If C=A then some others may feel that A is overdoing it. In any case, an anonymous robot will help to avoid all these sociological issues. – user121799 Jul 12 at 6:04
  • @marmot If B gets mad maybe they didn't copy or they feel they didn't copy. Don't think everyone acts mala fides. – CarLaTeX Jul 12 at 6:53
  • Certainly not everything but yes I had conversations in which first the user claimed it was not copied, but then there was a command defined in the original post that the plagiator copied but did not even use. But yes, there is the possibility of accidental similarity. One more reason to employ a robot, which finds the related post and gives the user the chance to explain the similarity without all that hassle. – user121799 Jul 12 at 7:01
  • While I do agree on the practical matters (that giving attribution for longstanding "wisdom" is quite hard), there is also a site license (CC-BY-SA) which requires appropriate licensing of posts, otherwise you infringe site copyright rules. If the user whom is copied from does not state otherwise, his posts are governed by this license… – TeXnician Jul 12 at 9:02
  • @TeXnician Yes, I know, I wrote it's a rule of the site, but sometime is not so neat to say if a post if copied or not. For example, if I implement an idea in TikZ and another user implement the same idea in pstricks, in my opinion they are completely different answers. – CarLaTeX Jul 12 at 9:19
  • Just for the records: if someone takes a working solution in PSTricks and the only thing they add is to translate it to TikZ, or the other way around, I do think that the original post is to be mentioned. @TeXnician This is not about longstanding wisdom. It is about devising a new strategy, writing a new command or style that is novel. – user121799 Jul 12 at 11:28
8

I don't see this as being a valuable feature. Attaching a list of possible related (or duplicate) answers to an existing answer will clutter the answer space with some related, but also some unrelated information. Just like the question suggestions feature a list based on some algorithm that results in hits and missed, you'll have the same with answers and that's not necessary.

Instead, post a comment with a link and it will show up in the related/linked sidebar. Don't worry about people getting upset about you posting a comment. If you think there's just cause, or you feel things are related/similar/duplicate, then act that way. Moreover, cast your votes accordingly. Downvote if you feel the answer shows unethical behaviour or, more primitively, doesn't answer the question.

Can one make the software that identifies related questions also list related answers, and attach these lists to the respective answers?

Yes, one can. But, for completeness,

  1. Such a request would have to cover the entire network and it would therefore be a better fit on Meta.SE. I'd encourage you to ask it there and see if you get any traction.

  2. It won't add the value you perceive that isn't already covered through other means (like comments).

You must log in to answer this question.