The "language can control thought" idea is called linguistic determinism, and has been considered bogus by linguists for a while now, at least the strong form of it.
On the other hand, that is not to say that an authority can't use language (in the sense of diction) and linguistic censorship as a powerful vehicle to shape thought. Many examples of this exist. Here are some real-world examples
* Modern Hebrew to foster an Israeli identity.
* Sanskritization of Hindustani into an Hindu-Urdu distinction of identity.
* Heavy and active Internet censorship of diction in China to reinforce what topics are taboo for social discussion or treasonous, and remind that Big Brother is watching.
https://www.researchgate.net/publication/12165254_Language_s...
This study found that a child's language production abilities predicts performance on a reorientation task: simply - young humans tend to perform like rats until they learn to use prepositional phrases.
Ex. sets of 3 blocks of various colors are presented to subjects. Either the experimenter gives a 'control lecture' showing possible upcoming sets and they just list the colors, or they teach the subject a name for each of the possible sets - gibberish like 'flurback' and 'orshwalg'. Present a list of sets, spending the same amount of time 'discussing' each set with each set - either with or without using names.
I would predict the 'names' group would significantly outperform in long-term memory formation of which sets were shown in which order.
Changing the emotional context of a small number of words; moving a concept from a socially unacceptable word into a socially acceptable one, seems fairly different. So I think we're talking about two different things here.
So while Orwell's "new speak" seems clearly to embody linguistic determinism, using "collateral damage" for "civilian deaths" inside a larger language may not be.