I'm writing this article with a preconception that most people reading will have an awareness of what a skinnerbox is and how it relates to gaming. If you do not understand this, please watch the following Extra Credits video in advance:
I'll say this outright - you can't have a game which capitalizes on dopamine rushes which sustains atmosphere. Why? Because the compulsion which makes someone want to jump right back into a game will always kick in long before the creative imagination of the player has rested enough to become creatively-invested in the world of the game again. Don't believe me? Try playing Diablo 1 sometime. I'm sorry to return to this one game time and time again, but it's a good workhorse for looking at these problems.
Early on in the game, you will get in a habit pattern of collecting loot, taking it back to the town to sell, upgrading your weapons and armor so you can more effectively go back out there and kill more demons (and collect more loot - you can see the skinnerbox loop here already). Now, this is all fine until you get to the point at which the pattern ceases to progress the story - when the kill-collect-sell-buy-kill loop ceases to directly feel aligned with the story and the game-world. For me this sets in around level 11, when there's been little-to-no story for a while, you realise that the townspeople just rabbit on about the same inane things time and weary time again, and the action starts to be decoupled from the storyline. You start to wonder, why am I treading this dull path?
The standard 'game developer' mindset to this problem is to add more bling - more spectacle and engagement to make the player care about the gameplay. This is a process of diminishing returns. The correct path is backwards - away from the gameplay and towards the story. Make the player see a reason to progress that involves the metaworld, not the individual achievements or the individual dopamine rushes that come from collecting more gear (when taken out of gameworld context after a while that loop starts to resemble Progress Quest).
The second and most important part of fixing the problem, is to remove the skinnerbox loop. While it's understandable that game developers want their clients to spend a lot of time in their gameworld (particularly for free-to-play models), if your aim is atmosphere, this approach will /never/ work. As a player who engages in the atmosphere of a game, I /want/ to be imaginatively immersed in that experience - for escapism, for thrill-seeking, for a myriad of other reasons. It's not a conscious process, but a subconscious process you can have some awareness of, if you choose to. However the simple fact of the matter is that the creative imagination has a point of exhaustion - it cannot sustain 'feeling engagement' for too long before it loses it's edge.
In a sane game, the player will jump out at that point, go and do something else, maybe go to work or spend time with the kids, then come back to the game when their creative psyche wants to be invested in the game again. Unfortunately for far too many games this never happens. What happens instead is the brain remembers the dopamine rush from the skinnerbox loop (in this case, kill-collect-sell-buy-kill) and wants to return to that far sooner than the atmosphere-appreciating creative subconscious does. That's when a game starts to become lethargic, unenervating, rote-work, arithmatic, competitive and boorish - all in pursuit of this lazy skinnerbox design.
Now, I'm not against grunt-work or grinding. Early on in a game it can be kind of fun to have this sort of activity. Even the dopamine loop can be used positively to get the player initially invested in the game. But you have to get off that pony and onto a horse - I'd say just before halfway through the game. In terms of Diablo 1 that would've been best simulated by removing the amount of item drops, making what was dropped far more useful (not merely sellable) and realistic (bats dont carry swords), and by changing the game-pattern to "invest yourself in creatively-fighting your way through these monsters so you can get to an [item cache]/[point which progresses the plot]/[area which changes the skills and stats of the player]".
And you can't rely on the player to change this for themselves (for example, in the case of Diablo, simply not picking up so much stuff), because by this point in the game, the habit pattern becomes entrenched. Otherwise it just becomes a grind, for the entire game. And there's nothing that dulls the creative imagination more than a continual grind.
Games which suck at this:
Diablo 2 (AAA, payware): The sequel took the kill-collect-sell-buy-kill loop of the first game and amplified it 100-fold. Item drops were constant and ubiquitous, variation of equipment was far more vast, making a meta-game of choosing which items to even /look/ at a necessity if you wanted to get any enjoyment out of the game. The addictive quality of this loop far overrode whatever creative investiture could be had in the game and the game's plot, and spoiled any sense of atmosphere. Luckily the game also ruined it's atmosphere via a number of other gameplay quadrants, which I might go into later - so it was no great loss.
WoW/Everquest/MMORPG's in general (AAA, variable): Skinnerbox tactics are a standard way to keep a player involved and paying/playing in subscription-based online roleplayer games. Free-to-play is disrupting this model somewhat, but the meme remains.
Games which are very good at this:
Psychonauts (AAA, payware) : Though there were item hunts, beat-em-up sequences and collecting-quests in the game, they were always inherent to the gameworld - they not only served a purpose within the world of the game, but what you were doing made sense within the context of the game. And the process you had to go through (particularly for the item hunts) was always creative, cognitively-challenging, and unique for each item, making it less of a chore and more of an adventure in-and-of-itself. Though there was a particularly hard meta-game of collecting all of the 'figments' (2-dimensional mental artifacts inside people's heads), this was not necessary to complete the game, while it had it's own in-game rewards. And the gameworld itself was so full of atmosphere, uniqueness and fun that I chose to go through and find all of these - simply to spend more time in the gameworld. Now that's clever game design.
Star Control 2 (AA, originally payware, now freeware): Although the early third of the game had a fair bit of grinding in terms of hunting/gathering for minerals and upgrading your ship to better fight things, this was balanced with an incredible story, great wit and dialogue, and by halfway through the game, you no longer needed to do much in the way of grinding because the progression pushed you into more exploration and dialogue with other alien races, rather than resource gathering. If I had a criticism it would be that the grinding is never explicitly made redundant until the end; ideally this should've happened much earlier to progress the plot-aspect of the game and to enable more star exploration.