[In this in-depth Gamasutra analysis, game professional Breslin examines the history and current state of the 'sandbox game', looking at modern games like Grand Theft Auto IV and Spore to see how they fulfill the concept of unlimited, unfettered creativity.]
It's such a buzzword nowadays -- sandbox. It is a very abstract concept, like "liberty" or "love," so there are a lot of varieties. Modern sandbox games draw from a wide range of design structures: from open-world design to emergent behavior, from automation of believable agents to multi-threaded or non-linear story.
Being applied in such a broad range of situations means that the word "sandbox" risks its meaning becoming watered-down, confused, and sometimes forgotten.
On the other hand, it remains the single most important design issue for current and future generation games. It's a great blessing to spend a while in serious thought, on such a critically defining idea.
Let us take this opportunity, then, to reflect upon the different ideas we as gamers and designers have of the notion. Let us think about where the whole idea came from, and consider where we've come to, as we look toward the many futures of this wonderful little concept.
Just to get rolling, let's consider the concept in abstract:
The Ironies of the Metaphor
The concept of sandbox-style gameplay, as we know, suggests more-or-less undirected free-play. The metaphor is a child playing in a sandbox: the child produces a world from sand, the most basic of material. This in contrast to a game where the upper-level content is presented fully formed and ordered.
The metaphor of "sandbox" suggests something pure and free. It implies that it is a young child in the sandbox (and a pre-videogame child at that, with no toys), and assumes an idealized childhood imagination, an unlimited creativity. It is a good metaphor, and a useful one, but the metaphor is also a little misleading, insofar as it suggests a sort of dream-world imaginative capability of the audience, which is not always justified.
The implications of the metaphor are not necessarily carried over by game designers: we anticipate that less imaginative players will get less out of a sandbox game, and this is fine. But even so, the idea of leveraging the player's imagination is quite ambitious, and more than a little risky.
By itself, this design concept is so ambitious -- "give them a sandbox, and they will build castles" -- that it must be met with a far greater investment in making the sandbox actually work, which generally means much more money and time invested across all levels of production, and particularly upper-level design and writing. It means an especially close relationship between programming and upper-level design, with the anticipation that the upper-level design will often determine the lower-level specifications.
If normal game design is developing upper-level material (missions, etc.) based on an engine, sandbox design is writing an engine to express upper-level gameplay concepts. Of course, it's very silly to put it so simply as that: indeed, game production is always a back-and-forth between programming and design.
But basically, sandbox design requires the development of engines which enable open exploration in various ways, engines which support upper-level sandbox design by providing systems for the handling of the sandbox elements. It's fun work for a systems programmer, but it's not easy. Then, on the upper level, sandbox requires design which emphasizes and encourages free-play, as well as the development and implementation of a wide range of dynamic interactive elements.
And it's not just the presentation of the sandbox elements and the play space, but it's handling all of the player's various interactions, all the possible combinations. To say it very simply, a typical game must respond to correct input, while a sandbox game must reward all input.
While a physical sandbox is very easy to build (compared with most other toys), the sandbox game generally requires far more work than similar, less open and responsive games. Unlike the person who builds a physical sandbox, a sandbox-style game designer cannot simply offload the creative effort onto the gamer.
The Necessary Framework
Unfortunately, sandbox design is sometimes taken to justify the exact opposite: "sandbox" sometimes serves as an excuse for less investment, particularly in high-level design. Sandbox elements can be mistakenly taken as fair replacements of narrative content; indeed, many games have missed their potential because they imagined that free-play would compensate for a lack of narrative. But even for our idealized child, playing around in a physical sandbox gets old pretty quick.
This principle design problem of sandbox-oriented gameplay is already subtly suggested by the sandbox metaphor itself: a child playing in a sandbox needs a lot of direction if they're going to have very much fun. They need toys first, and they need to be given ideas of things they can do with them. The parent needs to provide a meaningful framework. Just dropping a kid in a sandbox does not work.
The same is true of sandbox design. If the design effort fails to produce a game rich in intriguing potential, it's very much like shipping a literal sandbox. -- Imagine a game-box literally filled with sand: the open-minded player might enjoy playing in the sand a bit, but the gameplay really isn't worth a lot.
The necessary framework guides the presentation of the sandbox elements as the world develops and unfolds. This is often expressed as a reward system, which can involve new areas to explore and new stuff to do, more difficult gameplay structures to navigate, more story unfolding, more missions becoming available, and so on. It can be based on exploring the space of the game (exploring Liberty City for example), and it can be based on watching the game-world develop over time.
It often means scattering a great number of narrative elements across the game-world for the player to discover. Rather than presenting the sandbox as "here's a box of toys, goodbye," the framework gives some strategic order to the game's elements, a presentational structure -- and thus it gives the player periodic rewards for playing.
A common misconception is that the stories of sandbox games are not determined by the game's developer: instead, the story is supposedly determined and directed by the player. But even designers of the most free-form sandbox games must specialize in producing worlds which are geared towards making that free-play fun. If the sandbox is interesting (and this is by no means guaranteed!), then the game's potentialities, the potential interest and fun -- including the narrative undercurrent and whatever else makes the free-play engaging and worth the time -- are all very carefully handled by the developer.
"Sandbox" sometimes challenges traditional narrative, but it always puts something new in its place. -- Thus, it does not remove the narrative, but rather transforms predetermined narrative into dynamic, responsive narrative. In other words, the sandbox game distinguished itself by making the responses more significant and meaningful.
This is perhaps contrary to an image of sandbox play which emphasizes pure freedom, but again, sand by itself is not much fun. Automated, complex, and perhaps most of all, directed responsiveness is essential to sandbox play, and the more complex and responsive the world, the more interesting the sandbox.
What is interesting about the sandbox form is not that it allows full freedom, but that it generalizes and parameterizes, it finds arenas for agency and gently crafts the potential space of the game. It fosters a sense of free-play and exploration of that space. It engenders a sense of player control, without actually handing over the reins entirely.
Prelude: Leisurely Play, Discovering Elite
As popular gaming moved from the arcade into the living room, the stage was set for a less frantic, more leisurely presentation of the action -- gaming "off the clock."
No longer was there a need to kill the player off as quickly as possible, in order that he (or the people waiting in line behind him) insert another quarter.
It became feasible and even desirable to give the player a break from the action, some time to thoughtfully explore his environment, rather than race him towards inevitable destruction.
Thus were born a number of popular gaming genres, including the adventure game and the flight simulator.
The adventure game popularly began in 1978 with Warren Robinett's breakthrough game for the Atari 2600, entitled, appropriately enough, Adventure. 1980 saw the release of the more sophisticated and seminal Rogue and Ultima.
This genre moved through the popular classic Pitfall and the unforgettable catastrophe E.T. (both, 1982). The player now had his leisure -- his world was now basically open. Pac-Man had broken out of his labyrinth -- and found a real world waiting for him outside.
A seemingly distant genre, the flight-simulator, popularly began in 1980 with subLOGIC's appropriately-titled Flight Simulator, later licensed to Microsoft. These games were groundbreaking for the sense of freedom -- and what better sense of freedom than flying through the open air? Their basic problem was that the air was empty, so there was literally everywhere to go and nothing to do. The game was the movement alone.
Movement alone is a fantastic concept for gaming, of course, as has been well demonstrated by recent parkour-inspired games (Assassin's Creed, Mirror's Edge), and the closely-related genre of sports games. Also, exploration is the fundamental gameplay concept of open-world games. So, even while the early flight simulators were devoid of narrative or action, they were perhaps the first pure expression of open-world joy.
Then there came Elite (1983), which synthesized these emerging forces, and in so doing shifted the paradigm.
Elite was outstanding in many ways. Its graphics engine was original and groundbreaking: wireframe 3D graphics with hidden-line removal was a big deal back then. The auto-generation of the universe was brilliant, and its combat was clever (although dogfighting had already been incorporated by subLOGIC/Microsoft's Flight Simulator back in 1982). Its economy was a game just by itself, and it had a rich gameplay all around.
But Elite was truly profound because it presented a game-world space and a freedom of movement and choice that for the first time felt real and unbounded. The game-world no longer appeared to be a closed labyrinth or a hilly continuum, but was now an open universe -- and so the game-world metaphor began to operate on a new level.
With The Seven Cities of Gold (1984), this was of course the birth of a genre: the trade/exploration/combat/adventure sandbox, typically in space or at sea (key metaphors of freedom). The successors are far too numerous to list, but they include: Starflight (1986), Pirates! (1987), Star Control (1990), Privateer (1993, and following), X (1999, and following), Freelancer (2003), Darkstar One (2006), SpaceForce 2 (2007).
In the whole history of computer games, there have been only two other innovations which are on the same level as this moment: 1) the explosion of multi-player; and 2) the paradigm-shift from 2D "platform" to 3D world -- the latter already anticipated by Elite's cockpit view, though this was already done in the popular arena by the arcade game Battlezone (1980).
(Technically speaking, Jim Bowery's 1974 game Spasim was the first multiplayer 3D combat, but as it ran on a PLATO network mainframe, its audience was relatively small and specialized.)
However, it would be about sixteen years before game designers began to use the term "sandbox" to describe this kind of free-form play. Nevertheless, the concept of the open game-world is essentially the same, from Elite all the way to Assassin's Creed, Spore, or GTAIV.
The intervening years saw many trends in free-play, the most popular of which was the city-building game. It began in 1982 with Utopia, but the city-building genre really came into its own when it ceased to be strategic/competitive and became instead an exercise in "free" building for its own sake.
The genre grew out of the natural pleasure of designing game-worlds -- a pleasure that game developers experience all the time. One developer, Will Wright, thought that it would be a good idea to share this joy as directly as possible, and this insight led to the development of SimCity, which became a record-breaking success, defining one of the largest genres of the 1990s.
Sometimes this sort of free play was blended with economic simulation, in such as the Tycoon games, starting with Railroad Tycoon (1990). Various more-or-less competition- and objective-oriented games joined its ranks throughout the following decade, from SimIsle to Capitalism (both 1995).
Opening game design to the player, even to a very limited degree, heralded modern player-generated-content games, from Second Life to LittleBigPlanet to Spore.
Encouraging Player Experimentation
The metaphor of the "sandbox game" finally emerged at the turn of the century, around the publication The Sims and the following year, Grand Theft Auto III, the two games which are traditionally considered the two original and canonical "sandbox" games.
The invention of the term did indeed accompany a new development in game design, but this was not, as the term suggests, player freedom, which was already available by any number of means: non-linearity; the lack of objectives or central storyline; automatic variation of the game-world and game-behavior.
It was in terms of responsiveness and encouraging player experimentation that these games represented a gradual but transformative change in game design.
"Sandbox" was a new development because it indicated a new promise: automated responsiveness to player behavior. In this sense it does not mean "free play," "non-linear," and the rest; rather, it indicates that which makes this style of play specifically and particularly interesting in its own right.
Most of all, this meant a radical development in design detail. The evolution of sandbox-oriented quality between GTA2 and GTAIII is truly astounding. The switch from bird's eye to 3D opened the world and shifted it from cartoonish "Hot Wheels" platformer to a realistic city. But the critical part was that the writing and detail followed through on the promise.
As mentioned at the beginning, sandbox design facilitates and encourages a sense of player freedom, while providing a framework for play and a rich and detailed world for interaction. This was definitively achieved by The Sims in 2000, and in 2001, Grand Theft Auto III. Let's now consider their innovations, starting with The Sims.
Towards Believable Characters: Psychological Games, A-Life, and AI
The amazing commercial and cultural success of The Sims might suggest that it was entirely new -- which means we are likely to forget that the genre began with Little Computer People (1985), even though the latter "game" was well celebrated in its time.
This is the birth of the mind game, the virtual seduction.
There were several studies in the 1990s, of what gameplay structures and presentation/interface regimes increase attachment, what exploits the player's tendency and desire to interact in a seemingly meaningful way with the artificial character. These always somehow literalize the metaphor of the game-world, bring the player into the virtual space and enmesh him there: enabling "physical" contact (mouse-petting), sharing "space" (e.g., the player and the character can manipulate the same on-screen objects), and so on.
Today we are so close to such virtuality that it has perhaps become difficult to observe its mechanism, but a primary aspect of sandbox play is the formation of a psychological illusion of contiguity, if not continuity.
Thinking more towards psychologically effective programming, let us consider the dual nature of AI. As any AI designer or programmer will tell you, the task of designing a "believable NPC" involves fostering an appearance or impression of that elusive philosophical notion of intelligence: the psychological impression of intelligence.
What contributes to this impression, however -- the underlying program -- is more or less "intelligent" in an entirely different sense. Where we speak of the "intelligence" of the program, we mean only the level of autonomy and generality. This is not the place to get into the specific maneuvers and techniques, but be assured that relatively simple programming can lead to really convincing NPC AI, and really it's mostly in the presentation.
The NPC programmer's plan, then, is essentially to write suggestive and interpretable behavior, so that the player will "read in" a lot more sophistication than is actually present. Computer-players can be good at winning a chess match or a combat, which has relatively simple rules, easily-validated success cases, etc.
But beyond this, the question of truly intelligent programming (in that ephemeral, philosophical and psychological sense) is well beyond our technological horizon and may well remain there forever. The question of NPC personality in games is always the question of faking it.
The main reason that this trend towards believable characters is compelling for sandbox play is that the characters are, at bottom, more dynamic and interactable. They help "sell" the game world because they seem more realistic. Not "realistic" in the sense that they can ever hope to pass the Turing test, but realistic enough that they'll lull you into forgetting about their artificiality. The more intelligently the NPCs respond, the more the game feels like a free and open world.
AI is widely various and can be complicated, but in general it is effect-oriented. The programmer has in mind a goal behavior, and writes code to meet this objective. In comparison with AI, Artificial Life is bottom-up programming, and it's all about emergence. The emergent behavior is not necessarily even known in advance.
The Sims, and especially the range of games it inspired, was heavily influenced by technological developments in computer science during the 1990s, and in particular Alife. By 2000, this has developed into the art of manipulating automated NPC behavior, even in an otherwise traditional title, as we have for example in Majesty.
In this classic, there is no player-character, and little if any direct action by the player. Instead we have NPC agents whose behavior cannot be directly controlled, but only indirectly influenced in some way: add stimuli and enjoy watching how the automatons respond. It's a delightful gameplay model, which we look forward to revisiting in the forthcoming Majesty sequel.
Playing with automated systems, watching NPC AI agents interact with each other according to their program, or even watching Alife virtual organisms go about their daily life, has long been and remains a key sub-genre of sandbox play. Further, believable and self-motivated characters have become key to sandbox play, because they produce a rich space for interactivity and greatly help establish the open-world aesthetic. But in another style of sandbox games, the game space itself plays this role....
A Realistic World: Emergence, Robust Simulation
An emergent behavior is a consequence of the rules. Take the rules of, say, chess: the rules of chess do not explicitly refer to the concept of initiative or that opposite colored bishops tend to be drawish. But these and many other characteristics of the game are determined by the rules. We see emergent behavior in many complex physical systems (fluid mechanics in physics, for instance) -- or more to the point, we see it in the material happenings of any complex game world.
The various characteristics of explosive barrels in Doom is one canonical example. The rules which govern their behavior are very simple; nowhere does the program say anything about how they can be lined up for a chain reaction.
Once barrels started exploding in chain reaction, the virtual world had suddenly become robust, palpable, realistic. This was an amazing moment, but very little of the player's energy was invested in playing with the system -- yet. Doom struck a highly linear and simple tone, for it did nothing to encourage the player to experiment with the scenario.
From a certain point of view, Doom could be considered a sandbox: we remove the "EXIT" and the player wanders around killing baddies, doing as he likes. From the same point of view -- and this bears especially on how we commonly use the term nowadays -- the production of a "sandbox" game is a subtractive operation: subtract the missions, the main campaign, the narrative or whatever formatively binds the game's progression, and you have a "sandbox." The player can fool around without doing anything "on task" or so.
This is the sandbox we mean when we speak of "Sandbox Mode" (as opposed to "Campaign Mode"), and it is closely similar to how the term is used in software development.
In general terms, if one removed the objectives of a game to produce unguided play, or lack of narrative, one would makes a sandbox in some subtractive sense -- but not in a productive sense. True sandbox design means adding game behaviors which, in combination, produce interesting emergent behavior, but it also means adding some reward for free play. Emergence is good, but a free-play oriented framework is also necessary.
Metaplay and the Multiplayer Arena
While meta-play and multiplayer are certainly two entirely different phenomena, they have some things in common and they often happen simultaneously, so we might consider them loosely together.
Meta-play normally means a different approach to playing, where the player is no longer playing the game as it was designed, but messing around with it and doing amusing things. This includes exploring glitches, testing the game's limits, creating and pursuing personal objectives, and other things which were not necessarily intended by the game's designers. "Hmm, I wonder how many resource harvesters I could build..." or "How far can I drive off the track, and what happens then?" or "Can I finish the game without getting the spider ball?" -- this kind of play.
This relates back to our opening discussion of adventure games, whose design tends to be in the form of lock-and-key puzzles. One implicit challenge in such games, and one way by which mastery can be measured, is in figuring out the shortest route.
When the game is played in this attitude, the metaphor of adventure falls away, and the player instead thinks consciously of the underlying system, how to optimize given the rules of the system -- and even how to break the rules of the system. Though it operates on a different level, sequence-breaking is very sandboxy and very meta, and lock-and-key style adventure design encourages it, from Super Metroid (1994) to Switchball (2007).
The key here is that the game might support sandbox-style playfulness or meta-play, whether or not it was designed to do so. Sandbox is a much wider genre in terms of play than it is in terms of explicit design: a wide variety of games can be played in a sandbox style -- it just depends on the ingenuity and creativity of the player.
Even chess can be considered a sandbox game, if you look at it in the right way. It need not even be played as a competition: instead, you and your opponent could cooperatively explore the potentialities of the game, to see how certain interesting structures can emerge -- to "meta-play" the game, not competitively, but critically, analytically, imaginatively. (One could well argue that if you look at chess in the right way, meta-play happens quite frequently over the course of a normal competitive game.)
Indeed, any sufficiently complex game can be considered a sandbox if one of the aims of the players is to explore the implications of the game's rules. The metaphor of "game world" becomes strained, but it is possible to liken the space of potentiality opened by the rules to a game world, which the players can freely explore.
The point is that it does not take two opponents to play chess; instead, one can play in a creative way -- solving the eight queens problem, for instance, or producing an elegant endgame. The traditional card game solitaire is not really a sandbox game; but a solitary game of chess can be. The interesting point here is that there is a space of free-play potential even before the opponent enters the scene.
The case is similar with multiplayer: the game need not be specially designed to support rich sandbox gameplay; it needs no carefully-crafted narrative framework, no believable characters, and so on. By contrast, it takes only a modest arena to produce all the necessary strategic interest to support a rich multiplayer experience. Even the simplest of MUDs can do it. When it comes to multiplayer, we can strip things down quite a bit, as the opponent provides much of the necessary framework.
This is no argument against complex multiplayer worlds. World complexity often leads to more nuanced strategy, which is a good thing. But speaking minimally, all a multiplayer arena really needs is a set of rules.
Likewise, if the player approaches the game in either an ironic, analytical, or deeply-invested way, then the experience can rest on the simplest of pleasures -- such as riding horses around together in a wilderness, and looking at a randomly-generated landscape.
But on the other hand, if we subtract multi-player, or we subtract that meta- level of player interest, even the most realistic game-world can lose its interest very quickly. A realistic simulation can be a great multiplayer arena, and a great foundation for building a game-space, and it may indeed be fun to explore for a little while. But it must be recognized that realistic simulation on the one hand and gameplay/presentation on the other are very different phases in development, and if the principle challenge and purpose is not being supplied by multiplayer, then some directing framework must be supplied by the designer.
User Generated Content
Game design itself is, undoubtedly, the ultimate sandbox game: you the designer get to determine the game's objectives, and not only that, but also create and assemble the artwork and other presentation elements, balance the game as yousee fit -- create a whole world to play in.
Modding is quite similar to game design in that sense. The main difference is that modders do not write game engines and they do not design the larger framework. Their role tends to be limited to top-level design, though of course this varies from game to game.
Ten years ago, one would be wise to remark that "the future of gaming is modding." But over the course of the past decade, modding itself has become increasingly part of playing the game, and the line between playing and modding is now and forever blurred. From the simplest "scenario editors" of the late 1990s through Neverwinter Nights modding tools, to Crytek's Sandbox, game production has increasingly focused upon in enabling and encouraging player design, and today's games often present certain forms of design as a core ingredient of the gameplay.
For years, modders have been using Maya rather than Creature Creator. Spore's obvious innovation is that now every player gets to mod. Spore is not alone in this, of course. Indeed, LittleBigPlanet is arguably even more progressive in this area. The key is the creation of a game that works towards the objective of player-generated content, and designing tools to enable novice modeling and design.
We will be discussing Spore as a special case study, below. The salient point here is that while Spore makes an art of erasing the difference between modding and playing, the same thing has been done for years in a less integrated and novice-friendly manner, but far more completely.
Second Life is still another canonical example of a game that is designed to be modded. It produces an amazingly real analogy between (on the one hand) the clothing designer, construction worker, or architect/engineer, and (on the other hand) the 3D modeler. Players are sometimes even paid for their 3D models in real money, just as real-life carpenters are paid for their cabinetry work.
Whether people play perfected visions of themselves or ironical caricatures, the combination of multiplayer and modding assures a permanent place for Second Life. On the other hand, one key element is that there is no game-worthy interest. It is a pure sandbox, and so it suffers a lack of interest, from a lack of what we have been calling 'framework': a lack of direction.
Contemporary Case Studies
One of the best ways to evaluate the state-of-the-art of Sandbox design is to consider modern expressions of the design. General theories about this design strategy may help guide future development efforts, but practical analysis will explore how these ideas are playing out in today's game studio.
And seeing what's going right and what's problematic in practical terms can be a big help for guiding future development, and expanding the theory of sandbox design.
For our first example, we'll take:
One of the main stumbling blocks for this title is the dynamic nature of the game world, and the manner in which scenarios are spawned. The scenarios themselves carve out only a very small area, and are immediately surrounded by random "sandbox-world" elements. This was a relatively early programming and presentation decision, which determined that there would be little design control over the extended environment of a given mission.
Not to doubt the efforts of the design team, of course. These kinds of questions are extremely difficult and require an enormous amount of work. When you're innovating, it's an uphill negotiation all the way. But it can still be said that the design, while it brings "sandbox" design into a new dimension, also prohibits the kind of finely-tuned level design of conventional sneaker games. As we have already suggested, sandbox carries both promise and problem. In this case, the main problem is a lack of upper-level development.
Even the principal fan-made FAQ for the game complains about its gameplay.
"Once you experience the same old objectives over and over, then you go crazy and just want to get to the end." The problem is that the gameplay is underdeveloped or barebones. The thwarted hope is that the design and mission objectives would become more sophisticated.
Chris Kohler of Wired's Game|Life blog also found the game lacking in terms of design-depth.
It's not a perfect review: Kohler does not recognize how well Assassin's Creed rewards clever and patient play. But on the other hand, the game's subtlety will be unappreciated by many other players as well.
The game should probably have spent more time explaining/training the correct way to play, through more nuanced training missions early on. And then the game could have reinforced and telegraphed this message by more finely-tuned missions throughout the campaign.
Even something simple like adding a Hitman-inspired "Silent Assassin" reward for particularly delicate play (at critical mission junctures) -- this would have gone a long way towards communicating the concept of excellent play, and rewarding clever execution.
Thse design considerations would not fix the main problem, however. The principle dilemma is the flat, repetitive mission offering, combined with the shallow game-world depth. It is commonplace for a sandbox game to require the player to perform a few stock random-generated missions between each customized major campaign, but the variation and range of random missions are insufficient. There is not nearly enough range and variation to suppress their repetitive nature.
The world would be far richer and more like a sandbox world if there were twenty well-worked-out characters per city, who the player could visit to talk and receive missions -- and play off each other. Allowing the player to work with different warring factions would greatly add to the sense of immersion.
Inevitably, scenarios of greater sophistication and wider variation are needed throughout. Random scenarios are good, but over-use of default scenarios means repetition. Ideally, each scenario is customized. Players allow recycling, but they don't like excess repetition and greatly appreciate custom scenarios.
There is no simple solution; it's the kind of problem that has to get worked out over many months if not years. But the general solution is extremely simple: a much larger fraction of development investment must be spent on upper-level design (mission design and writing).
The unifying metaphor makes or breaks the game, for it is a game which works very hard to be viewed as greater than the sum of its parts. Some people will suspend disbelief, and feel a little joy as the overarching narrative emerges; and other people will see the seams.
Still others will look at it as "five games for the price of one," and then judge those five games harshly: many have observed they are "-lite" versions of other previous sandbox games, from Age of Empires to Masters of Orion.
And while it remains true that Spore explores many different genres of sandbox play -- it is almost a survey of the larger genre, from cellular A-life to empire building -- it is also true that the separate phases are not integrated, and each in its own area is simplistic and under-implemented.
If the game works, it is because of the larger aesthetic presentation, and the great risk the game takes is that it leaves its success or failure to the aesthetic sense of the player. Whether the game does indeed hang together -- whether a person buys into the idea of developing from a cell into a planet-jumper -- this is mostly subjective. And this lays bare a common challenge in sandbox design: player commitment to open story.
The main interest, perhaps, is the creature creator (along with the building and vehicle creator). These segments of the game share with the audience a critical aspect of modern game development (3D modeling). In this, the design is highly consistent with one of Wright's major statements on the origin of his particular brand of sandbox design: that game design is so fun in itself that, if properly packaged, it can well be reinterpreted as gameplay itself. But let us consider another quite intriguing quotation:
Your heroic efforts have proven you deserving, worthy of advancement to the next level of your existence. The universe you inhabit is but one of many countless worlds, unseen but yet connected. Your creative efforts have not gone unnoticed. Indeed they have spilled into these other, unseen worlds, just as your world has been enriched by them.
Thus speaks the god of the galaxy, upon completion of the final objective. (Though of course this is not the end in the strong sense: the sandbox subsists after this, so you can continue to expand your empire indefinitely.)
These words of the divinity (playfully named "Steve") are quite clever, from a writer's perspective. The game text is doing what a writer might call "reinterpretation," or more simply, "changing the subject in a clever way."
Steve begins by proposing yet another level for the game: "the next level of your existence." So far there have been five "levels of existence": cell, animal, tribal, civil, and space. By the logic of the game, if you win the fifth mini-game -- the space game -- then you should get mini-game number six. The concept of a multiverse is then invoked: this would be level six!
Then, this multiverse is reinterpreted as the multiplayer function of Spore, which you've already been doing: "your creative efforts [...] have spilled into other worlds, just as your world has been enriched by them." It's a brilliant poetic turn, in the sense that the sixth level is the reality of multiplayer. It's also a cop-out, in the sense that the game is done providing new stuff to do.
The sandbox persists, but it is now repetitive with no hope of new content or additional gameplay interest. In most cases, gameplay is geared towards meeting an objective in order to enable a new feature of the game, but here it is just the joy of play itself, and again it comes down to player commitment to the purely open game.
Grand Theft Auto IV
Since GTAIII, what has made this series work so well is the depth of detail, partly in graphics but primarily in the robustness and variability of the game world. Characters are fully implemented, and missions vary widely.
The game world is further fleshed out by numerous concurrent and optional missions, so there is always a sense of player choice or determination, even though the main campaign is mostly linear and determined. In short, a lot of production has gone into top-level design, and it pays off big time.
There are a large number of situations which do not respond logically, but the player happily forgives this because the implementation is already so detailed. Players greatly appreciate the amount of work that goes into making a robust game world, and the depth and breadth more than makes up for the periodic lapses, the seams and traces of artificiality.
The major structure is equivalent to standard RPG: main-line campaign missions plus optional missions. This is an extremely common and an excellent format for sandbox gameplay: one central campaign (itself perhaps multi-threaded), plus a large number of side-missions. The sandbox all by itself is strong, although making your own fun can get boring after a little while. But crucially, the writing is thoroughgoing, so in addition to base sandbox play, there's always a choice between things to do.
Sandbox play is essentially amoral/non-moral, in the sense that real action is often governed by the hypothetical: "What happens if I run this guy over?" -- this is not a malicious thought; on the contrary, it is quite playful. But this generates a problem: how should the game respond? The GTA solution is to mock realistic response without actually enforcing it, and characterize the PC as logically as narrative ingenuity is capable of.
Grand Theft Auto IV
Still, until GTAIV, the PC personality was something of a narrative problem; as others have written, the hero was a bi-polar thug for whom nothing was truly out of character. Such a character is not terribly interesting: this is facile characterization from the perspective of the writing, and it's not particularly compelling in the sense that we can't relate to the character.
With GTAIV, however, the PC is framed very carefully, and the scarred warrior turned ironical and embittered anarchist justifies much better the peculiar range of action of a GTA hero. This goes to show how well a carefully worked out narrative can match the gameplay.
The problem of NPC response is the simple problem of AI. In the absence of strong AI, all the responses have to be hand-written, which means there are going to be some edge cases no matter how thoroughgoing the design. As suggested above, certain lines -- though they may be realistic or logical -- are nevertheless problematic and counter-productive, and best to ignore even if this does create a seam.
In any case, NPCs reactions always follow certain lines, and practically speaking, they cannot be extremely finely tuned to player behavior. They will fail to respond or respond incorrectly if the player's behavior does not fall into predefined categories. That is, quite simply, unavoidable.
It's a problem, but on a case-by-case basis it can be finessed by clever writing and design. For example, at the beginning of GTA IV, the "first drive" is bound to be messy, of course: the player is still learning the controls. In addition, the player needs to be eased into the peculiarities of believable world minus consequences, and NPCs who seem real, but who are strangely tolerant of behavior that would be taken as extreme in the real world. We need an NPC to guide the player, as part of the intro-tutorial, but how is the NPC going to respond if the player crashes the car?
In this case the writing supplies the solution: the NPC is extremely intoxicated, so he doesn't really notice (or doesn't really care) that the vehicle is getting trashed. The NPC is there to say lightly-disguised tutorial-oriented instruction, and to introduce the player to a world without consequences -- all under cover of the intoxication conceit.
This kind of case-by-case problem solving is the stuff of strong design. GTA's implicit recommendation is that the game be designed very carefully and strategically, with such problems well in mind, so as to avoid as many problems as possible.
Beyond that, where characters are concerned, it is probably best to treat the remaining problems by brute force: if there are ten edge cases in this scenario, script responses for all ten. Solving the general AI problem sufficiently is so far off that it's not even worth thinking about. Even the efforts of avant-garde artists (such as http://www.interactivestory.net/) all demonstrate that the bottleneck is case-handling and writing. In short, the bottleneck for apparently-intelligent behavior is writing for each case.
In direct opposition to the notion implied by the metaphor, sandbox design requires more top-level design, not less. To be frank: the great risk of the sandbox is that it can be boring.
Where the sandbox is the main part of the game, sometimes the game is impossible to "master": instead you can just juke around with it until it grows tiresome. In this case, there's no artfully-crafted narrative, so no climax; there's no reward scheme or gameplay-building (such as technique-training and gradation of difficulty). All told, it is weak on conventional game-design fundamentals. The added freedom makes up for this to an extent, but the problems must still be addressed and overcome by the gameplay design itself.
The space-empire-building game X3 is a solid example of this. There is a main campaign, but it is secondary to the sandbox play. (It wasn't carefully designed or written, which is unfortunate, but let's leave that aside.) The game shines as a sandbox involving the handling of complex systems, and the emergent behaviors are interesting, on a systemic scale.
In the end, however, managing the empire becomes something like a full-time job, the goal and reward being that you have a stranglehold on a single-player universe. From one perspective, it's fun to gradually corner and master a dynamic system, but from another perspective it's -- well, let's just say that one might be inspired to existential or self-ironical thoughts: "why am I doing this?"
Where there is a narrative alongside the sandbox play, one of the main problems is the pacing of the narrative. Do we allow free play simultaneous with the progression of the story? If not, how and with what justification do we interrupt the story in order to allow periods of free play? And then, how do we avoid the scenario where a lengthy round of free-play strains the pacing of the story?
As always, the particular solution will depend on the situation, on the game's unique features and format. But to take an example, another space game, Freelancer (2003), handles this exceptionally well. Gameplay is divided into discrete chapters, and as naturally as time passes between events there are periods of free play; the writing very carefully makes natural these breaks in the action.
One way the writer answered the problem of pacing: the non-player characters are always impatient with the player character. So no matter how long the player took, the response is logical. More critically, when the climax comes up, and pacing becomes particularly important, there are no "freelancing" episodes between the missions. This preserves narrative integrity, but the sandbox "free play" quality of the game is bracketed during this sensitive section. It goes to show that story and sandbox are sometimes very much competing principles.
Freelancer is especially worth mentioning because after the story finishes, the game becomes boring very quickly. There are only a few randomly-generated missions, and by the end of the game, the player has probably already played all of them. There may or may not be new places to explore (depending on how extensively the player freelanced during the main campaign), but there's nothing truly new.
The point here is that there's something cathartic about seeing the credits roll. Gradually losing interest as the game becomes more and more obviously repetitive -- this is an ending with a wimper. But on the other hand, there's always mods (and if you're technically inclined, there's always modding, which can be even more fun). And of course, there's always multiplayer.
The future is bright. The sandbox lives on in perpetuity. And life is good.