I think these types of questions have to do with catching up on your knowledge of modern chess engines and the humanity of them. We're going to assume these are engines that are playing at the top of their strength.
Quote:
I assume that when choosing between two moves they will absolutely never choose one that they discern is a forced checkmate against them regardless of the unlikelihood that a human opponent will find it...
That is correct. There may be some positions
the engine won't quite understand , but it's rare for sure.
Quote:
...as long as the alternative does not result in that forced checkmate as far as they can see.
If it's choosing between a move that leads to mate in 7 and a move that leads to mate in 8, it depends on the settings of the engine but they'll still play the "best" move for their position. A move that leads to mate in 8 is better than a move that leads to mate in 7.
Quote:
(I'm guessing that human grandmasters might deviate from this algorithm, hoping the opponent will miss the checkmate, because the other move sucks. Correct me if I am wrong.)
Human players, especially strong ones, will most definitely play a complicated losing position than a simple winning one. You should never play "hope chess", though. (IMs and GMs will even play a completely lost position for stalemate tricks, sometimes.)
Quote:
While there is usually a theoretically "correct" gto move that the computer is usually finding, what about when it is already in the midst of a situation where it knows that a perfect opponent will checkmate it. If it is asked to play on is there still some sort of gto type criteria in the hope that the opponent messes up? Or in this situation, do the various computers have major differences in how they choose their moves.
I really would not think of this in terms of GTO. GTO is not a concept in chess. Yes, engines can play the "optimal" move but it's not really a term used in the chess world. I don't know about it "knowing a perfect opponent will checkmate it" because it's a chess engine that's supposed to win. The major engines have differences in how they play their moves, yes. Sometimes engines will make "waiting moves" or moves that don't make a ton of sense on the surface to human players, and when analyzing or faced with one of those types of moves it's confusing and/or strange. A random king move, for example, when the position doesn't call for it. There have been breakthroughs in the last few years with neural network engines that play a more human-like style. I think you should read up on those, because that's essentially what you're asking for.
Leela played itself millions of times to get an understanding of chess.
Maia is an engine that was developed through only human-played games. Maybe read up on that?