That's interesting. Not sure how to best take advantage of this scenario when the human opponent is a peer.
Probably bad for learning though. Computers don't play using practical ideas, and of course never explain moves to you.
100 games of this with a GM partner (where you talk about the moves afterwards) might be worth a lot though.
I watched this video: https://www.youtube.com/watch?v=mp8y0BSRzf0 from kingscrusher on Youtube. In the video, he shows a suggestion from the computer that would be a stretch to think of for a human but then investigates why that would be strong. Maybe such scenarios may give new insights chess in general? In any case, could chess.com put this up as a new variant? If so, what would be a good way to bring it to their attention? It might be easier to arrange then games with a GM (not saying better). It would enable to see the effect at least.
I think there are opportunities to building computers in explaining moves. Even a simple statement like If Nb4 then Qb2 is strong for your opponent. This could well be off-topic.
I use an engine all time and look at the moves it suggests. But I have to come up with the ideas on my own. Knowing to play a single strong move is not good enough. If you can't follow through then you may actually damage your position very soon afterwards!
You could post it in the suggestion forum I guess
http://www.chess.com/forum/category/suggestions
Since engines find moves with the help of massive calculation, they're fundamentally unable to explain why a move may be good or bad. (Other than something very simplistic like you're weakening a square, which may or may not even be the point.)
There is potential for creating an engine that could explain, but none of the current engines can do this. I think this is a widely held belief which, IMO, will be the main obstacle to making your idea popular.
So it would be best to first have a computer that can explain moves/positions? As chess requires a subset of language, it might be doable to automise creating sentences as has been done similarily for trainstations. Maybe with a lowered search-depth, the suggestions by the computer have more meaning to the human in the game? Chess.coms computer level hard could be of help to players at my level for example. What would be the best thing to suggest chess.com in your opinion? My initial idea, or to (help) create an explaining computer or a combination of both? Maybe players should decide for themselves how deep the computer should search.
How about a game where human and computer alternate turns in a chess-game?
So say a human makes say the odd-numbered moves (1, 3, 5, ...) and the computer makes the evennumbered moves (2, 4, 6, ...). The computer would be equally strong for black and white.
I think it gives advantage in learning and might be a nice variation that might take need care as players might not be 'able' to play without this partner. Any ideas or references to earlier mentioning of this? (This might already have been investigated, I'm curious about what it would do to the game)