Gfyuki

Some months back I got into a situation where a paper with information that I considered important to prove a point was in a language I cannot understand. Since it was important but not crucial, I ended up not using the paper. Except for its abstract, which was in English, the entire paper was in this other language. It had graphs and images that I could understand without knowing the language.

At the time, I was tempted to use Google Translate in order to get the general idea of the paper (in more detail than what is given in the abstract), while not necessarily trusting on the specific information that the translation gave me (it can give faulty results at times). Since this would still leave me with some measure of uncertainty — there was no way for me to be sure exactly what it was saying — I thought it better to just let the paper go.

But say that in a future situation I find a crucial paper in a language I don't understand: if I use Google Translate, using the paper in a situation that does not require specific information but only it's general gist, would citing it be ethically acceptable?

(It's almost the same situation as this question, but I'm not concerned with translating the entire paper, for this would take time/resources that I simply don't have access to.)

I think the fact that the paper is in a language you don't understand is somewhat of a red herring. Sure, it will affect the specific steps you'd take to try to understand the paper, but the ethics of "to cite, or not to cite" it is no different from a paper you don't understand written in a language you speak. You're generally supposed to provide a good-faith overview of the literature, with citations as appropriate. The fact that one paper is significantly harder to understand doesn't really change that.

My view is that it behooves you to, as far as possible*, and regardless of language, at least understand at a high level what the full papers you cite contain, and the parts of papers you use on a detailed level. Google Translate will be good enough in some cases, but for other cases it's better used as a starting point. If you find yourself doubting the machine translation it's certainly best to use it as a starting point. You might want to try to clarify uncertain points with the authors, ask a colleague (who might speak the language) for help, get a professional translation of parts of the paper, or use a secondary citation.

*In some fields they even study ancient languages to understand original texts. That might be overkill...

Reiterating the main points (as I see it) of the other answers: if it's relevant, cite it... both to show your own awareness, and to acknowledge prior art, whether or not it exactly impinged on your own work.

That is, whether or not an individual can fully vett a piece of work, through difficulty with the ambient language or whatever, acknowledgement of its existence is very important, I think. Citing things is different from endorsement, and is different from a claim that one has fully checked all details, or even read the whole thing through. Just be clear, in the citation, what use you made of it, or did not.

Once again, straightforward-ness, honesty, are better guides much of the time (too bad not always... I know...) than stylistic prescriptions.

As an example to perhaps not cite: if you happened to read arguably crack-pot documents on arXiv (for math, for example) purporting (implausibly) to do amazing things in your specialty, I think you can justify not citing... if the author has insufficient credibility, and no one else with any credibility has vouched for it. But, yes, what the heck is "credibility"? :)

I would lean to including it. Of course you need to be careful about relying on something you can't completely read but this applies to everything (even just hard math). But I would lean to more cites than less since it helps future researchers.

It is not like you are endorsing everything or giving some huge gift to make a cite. You are trying to help people find info. They can still do some evaluation or even get into tangential ideas.

If you just need a datapoint (say transition temperature of a well known and noncontroversial substance), I think just taking the datapoint is fine without translation. Depending on how much more you need, you can try struggling through it with partial info, ask someone to look at it, use Google Translate, etc. If it is crucial, just pay for a translation and charge to your grant (this is not that hard, have done it).