The idea that player choices in computer games can have consequences generated by the game in reaction to those choices is not a new or outlandish concept. In a way, that's the core of the feedback loop behind all computer games.
A choice and its consequence are usually jammed as closely to each other as possible. In first-person shooters, things start to happen on-screen as soon as you pull the trigger. The action/result loop can be even shorter in fighting games.
What happens, though, when choices and consequences are separated in time? What's the effect when the player performs some action or actions, and there's a consequence that pops up unexpectedly (but plausibly) from what the player did in the past?
A jack-in-the-box is a trivial example. The "jack-in-the-box effect" of older tanks bursting explosively after their ammunition was hit has a similarly short-term action/result connection. But computer games can make the jack-in-the-box effect more surprising. When a game presents the consequence some time after the player's action, the experience is less "I made that happen just now" and more "Hey, this game remembered what I did!"
Doing more to let games appear to remember player actions over longer stretches of in-game time is something I'd like to see used more often.
This could go in a couple of ways (or both).
MEMORY AS HISTORY
One way is to let individual player actions be pretty trivial and pass without any special results, but respond to some preset level of accumulated related actions. Getting an achievement for shooting 500 opponents is an example of this, as is being granted access to previously gated content after raising "faction" with some in-game NPC organization. In this mode players usually know exactly what they're doing and what they'll get. And that works for conventional follow-the-rules games.
But wouldn't it be interesting not to reveal all the possible player actions that the game can observe and count, or the reactions of the game to certain combinations of accumulated player actions?
This might not be a good fit for conventional "you play it to beat it" designs -- players who enjoy those games will probably find surprises frustrating, rather than pleasant, and developers of such games generally don't like player creativity. Unexpected results for additive actions might be a very good fit, though, for a game where much of the pleasure is in the exploration of the gameworld and its internal systems.
MEMORY AS FLAGS
The other way of "remembering" player actions is simply to set a flag for specifically detectable individual actions, then test that flag sometime later and trigger a consequence if the flag is set. This approach is often seen in computer roleplaying games. In Bethesda's Fallout 3, for example, the game plays out in slightly different ways depending on whether you choose to detonate the warhead in Megaton. A somewhat more exotic example is the way that your choices for Commander Shepard in the first and second Mass Effect games, as preserved in your final savegame files, are reflected in minor options in the second and third installments if you let them start by reading the previous savegames.
This mode of modeling memory could also be enhanced. Games could take important choices early on and deliver very different gameplay later on based on those choices. This is rare, but a very good recent example is in The Witcher 2. Your choice for Geralt toward the end of Act One dictates which of the two mutually exclusive Act Twos you get to play. (Not everyone was a fan of the specifics of that, but I think the idea itself was worth trying.)
The important thing about consequences for one-off player choices is that developers almost always want to plant big flashing neon signs around it: "Look! Important Choice Here! This Will Have Consequences!" That's not always a bad thing. In a typically mechanics-driven game where it's considered wrong to ever let the player be confused about anything, signposting an important choice simply meets player expectations.
Not flagging such choices might be OK (at least sometimes) in a more exploratory game, though. Part of the fun of exploration is figuring things out. This is why puzzles are common in games where the developers want to encourage exploratory play.
So in a game of discovery, maybe discrete player actions that have later consequences (minor or major) don't always have to be signposted. (There does need to be an obvious connection between the choice and the consequence, though. If the game doesn't clearly explain that the consequence flows from a specific action by the player, then it just looks random. In that case there's no value in implementing this feature. Realizing the connection is what makes a delayed consequence particularly interesting.)
CHOICES WITH SURPRISING CONSEQUENCES
In both of these cases it's a good idea to be up-front with players that choices they make may sometimes have important effects later on in the game, and players won't always know when they're making such a choice. Developers should be honest about this so that prospective players who absolutely hate not being able to control all outcomes understand that this may not be a game they'll enjoy. If that's done properly, then letting some actions have unexpected (but plausible!) consequences later could be a lot of fun for players who do enjoy interesting surprises.
Overall, I'm very happy to see games like Proteus and SoundSelf being made, and I'm looking forward to seeing how they evolve.