Translation is costly, but the cost of translating a popular article may be smaller than the total costs of all people who would otherwise have to read it in a foreign language.

As a model, let's say that reading an English article by a person with English as a first language, costs 1 point of energy. Reading the same article by a person with English as a second language, may cost 2 to 10 points, depending on how good that person is in English. Translating the article to another language would be perhaps 100 or 1000 points. Assuming these numbers, once you have enough readers with a native language X, it becomes globally cheaper to provide an official translation to X in addition to the original English article.

In the other direction, it may be useful to provide an ability to write articles in X, and then translate the popular ones to English. This is probably even more useful, because writing in a foreign language is more costly than reading. On the other hand, the filtering of the good articles (worth translating) must be done in the original language X, so there must already be a strong community.

On the other hand, if there are more versions of the article, the discussion will be split. But again, to some people this reduces the costs of participating. (Perhaps the best rated comments could be translated by volunteers too?)

Perhaps this is a bias when thinking about languages: One language would be cheaper than many languages. But in a situation where we already have people speaking fluently in different languages, translation may be cheaper than using one language. Especially if we can have a good filter for translating only the best things.