Gamasutra: The Art & Business of Making Gamesspacer
View All     RSS
October 31, 2014
arrowPress Releases
October 31, 2014
PR Newswire
View All

If you enjoy reading this site, you might also want to check out these UBM Tech sites:

Warren Spector: Who forgot the 'role' in Role-Playing Games?
Warren Spector: Who forgot the 'role' in Role-Playing Games? Exclusive GDMag Exclusive
July 26, 2013 | By Warren Spector

July 26, 2013 | By Warren Spector
More: GD Mag, Design, GD Mag Exclusive

In this reprint from the September 1998 issue of Game Developer magazine, famed Deus Ex creator Warren Spector envisions a sandbox future where roleplaying takes center stage.

The oddest thing about computer role-playing games today is that you never hear anyone talk about the importance of playing a role. You hear about "400 character classes!" "6,753 unique skills!" "827 errand boy missions!" and "A world so big you won't want to explore it all!" Give it a rest. This is shallow. It's silly. It betrays our geeky roots in paper gaming (a medium with only a dangerous, superficial relation to electronic gaming).

Role-playing isn't about statistics or exploring randomly generated worlds of crate-filled buildings. It isn't about random quests and combat encounters every sixteen steps. It isn't even about +37 Swords of Instant Critical Hits that do Double Damage From Behind! Roleplaying is about giving players the freedom to act as they see fit, within the framework of a story we provide.

Role-playing is about characters developing in unique and meaningful ways as a result of player choices. It's about trying new behaviors in a safe setting before we try them in the real world. In the space I have here, I can't tell you how we make a game that allows us to do all that. But let's start by identifying problems, and by looking at character, setting, and story, and how we usually approach them.


Most RPGs define characters by an arbitrary "class" and/or a tiresome list of statistics. Characters typically have 6-12 attributes (strength, intelligence, and so on) and dozens of skills tracked at a fine level of granularity (lockpick score of 12, sharpshooter 72, computer hacker 53). Secret die rolls determine success or failure in skill use. The problem with this is that two players can do exactly the same thing and get different results because of insignificant differences between their characters. The difference between a 72 and a 73 shouldn't have any impact on game play. Does anyone think this is fun? We have to come up with game systems that tell players what their characters are capable of doing and why they succeeded (or failed). In a computer game, we don't need 42 skills tied to percentile die rolls to simulate skill use. We're clever. We can come up with something better. Leave the dice and character sheets to paper gamers.

RPGs often use characters' abilities to bake bread, charm NPCs, and so on. Yawn. Some think hack-and-slash is a more riveting way to use characters' attributes. Ah, combat! It's relatively easy to simulate and it gets adrenaline pumping. That's not enough. Here's a radical concept: let players control when and if combat happens. Our goal should be to make combat an option, but not always the best, and never the only one. Encourage noncombat interactions, especially conversation.

We can't compromise conversation -- a terrific tool for differentiating characters -- and still call a game an RPG. Here are some ideas for improving conversations and game play:
  • Conversations should reflect game state. Nothing's goofier than NPCs who keep talking while orcs hack them to bits.
  • Conversations should not involve lists of keywords. They're not fun, nor are they revealing of character. They're filler. They reduce conversations to the status of another stupid puzzle.
  • Conversations should reveal things about NPCs; your responses should reveal things about you. The best way to accomplish this is to make "Yes/No" options the rule in conversational interaction with NPCs. Take, for example, a situation in which you and a friendly NPC face several enemies. The friend says, "I'll hold them off while you escape and Do Important Things..." Leave, and your friend is doomed. Stay, and your mission may come to an end. A Yes/No decision becomes a dramatic moment that reveals something about your friend and about you. That's very compelling game play.
Conversations are made interesting by the things they reveal about the characters speaking, the game world, and the world's state -- not the number of branches in a conversation tree.


I've worked on games in which it takes hours to walk from one side of town to the other. Many popular, award-winning RPGs boast of hundreds of generic towns and randomly generated quests. The shallow simulation of huge environments isn't a good thing. Providing dialogue for scads of NPCs means none of them has anything interesting to say. Creating an entire country means any single building will be devoid of useful objects. It's a matter of time and storage space, and no amount of whack-on-the-side-of-the-head thinking allows you to finesse your way around the problems. Limit the size of your world. Provide several smaller maps. Increase the density of interaction. This accomplishes several goals:
  • Players can explore without searching for something exciting to do. Aimless wandering is the enemy of fun.
  • Developers can populate the world more densely with characters, objects, and quests, and give the illusion of a place with a life of its own.
  • Action can be tailored to player skill. Difficulty can be increased easily as players get deeper into the game.
  • Developers can create more varied locations than in a sprawling world. This last point is critical, and most RPGs do this well. However, most RPGs feature wacky environments straight out of designers' fevered imaginations. It's not asking too much to think in terms of believable, recognizable locations instead of arbitrary game spaces. We should try to acknowledge the conventions of the everyday, even when we create fantasy worlds. In the real world, you can tell you're in a bedroom, as opposed to a bathroom, the instant you enter because of size, placement, and furnishings. More game designers should realize this.
Some games do hint at the possibilities of believable environments, but they don't go far enough. In Duke Nukem (a game I loved), the environment was a gimmick. You knew you were in a movie theater, and you could switch the projector on and watch a bikini'ed babe do her thing -- let's talk about sexism another time -- but you couldn't switch on that projector and blind a sniper before he fired. Imagine if shooting a fire hydrant allowed you to douse a fire. The Ultima games go further, but not always in significant ways (mea culpa!) -- the key is not that every plate and knife and fork be usable, or that players reap wheat, grind it into flour, and bake it into bread. The key is recreating realistic locations and object interactions that are exciting. Give players believable worlds with lots of usable objects that produce predictable, useful results. Let them blast barricades, freeze enemies and then shatter them. Create worlds where water damages paper and gratings creak beneath players' feet.

Every game problem should have multiple solutions, by design or because alternatives arise naturally out of the simulation. How players deal with the problems they encounter (whether they choose violence over cleverness, talk first and shoot later, and so on) should affect subsequent interactions with the denizens of the game world as well as the substance of later missions.


Is it just me, or does it seem like every RPG drops players into a huge, all-but-empty world and says, "Go. Hope you find some fun."? Man, have I been guilty of that. After stumbling around for a couple hours, players may even find a clue that they're supposed to Kill the Evil Foozle. It's almost as if there's some unspoken rule against offering RPGers clear goals. The trick shouldn't be figuring out what you're supposed to do (which isn't much fun); the trick should be figuring out how to accomplish what you know you have to accomplish. New goals can be revealed as you go, but damn it, reveal those goals! And make those goals more compelling than "kill everything you see," okay? If working with Richard Garriott taught me anything (and, believe me, it did) it's that an RPG can be more than just a slugfest. More than any other medium of expression, gaming lets people find their own answers to tough questions, rather than imposing an artist's vision of the world on them. It doesn't matter what issues we explore -- tolerance, morality, relationships, whatever -- but let's explore something.

Dungeon crawls are all well and good, but we can allow players to explore who they are and what they actually believe. Unlike authors and filmmakers, we can give people the opportunity to test behaviors they'd never try in the real world. I feel we have an obligation to do that. If we provide only one answer (usually violent) we do our medium and our players a disservice.

Allow players to make choices and then show the ramifications of those choices: kill everything you see and suffer the consequences; play the pacifist and pay a different price. Games should be rife with ethical dilemmas rather than right and wrong choices. "What are you fighting for?" and "How do you achieve your goals?" should be unavoidable questions. When all is said and done, story goals and tough questions are just tools used to suck in players. If we create small, deep, object-rich simulations that allow multiple solutions to tough problems, players will inevitably stumble upon the "real" goal of an RPG -- to grow a unique alter ego.

Doing everything I've outlined above won't assure you a hit and accolades from peers, press, and players. All I know is we have to try. We have to fail gloriously. If we keep settling for RPGs that could have been made five or ten years ago -- and that describes every RPG released in the last couple of years -- we're doomed.

Related Jobs

Next Games
Next Games — Helsinki, Finland

Senior Level Designer
Magic Leap, Inc.
Magic Leap, Inc. — Wellington, New Zealand

Level Designer
Grover Gaming
Grover Gaming — Greenville, North Carolina, United States

3D Generalist / Artist
Demiurge Studios, Inc.
Demiurge Studios, Inc. — Cambridge, Massachusetts, United States

Lead System Designer


Rob Lockhart
profile image
I spend a lot of time thinking about how to make a video game that feels like improv. That's what 'role-playing' means to me. Giving a character details and relationships. When I create a character onstage, I don't think "What are this character's abilities?"

I can't say that this line of thinking has really gotten me anywhere yet, but I'll keep trying.

Aleksander Adamkiewicz
profile image
Mr. Spector is under the impression that arbitrary choices make for freedom and "fun" in a game. I wonder if he still holds this to be true to this day.

I couldn't be more opposed to this notion, if you can't "play" it, whats the point?

You will never have a meaningful conversation in a video game as Mr. Spector suggests, as long as we can't program an AI that is either fully sapient or can emulate sapience to a complexity close to our own.
Otherwise the player is always only choosing from a premade set of decisions provided by the developer, simulating this doesn't change it as the simulation will always have a limit in complexity and will never actually emulate a real conversation. As far as I know nothing passed the Turing Test yet.

Furthermore as long as this conversation does not lead to a winnable state or follows rules (those evil stat-based rules that Mr. Spector abhores) it is of no real consequence to the game. It would boil down to just choosing content you want to see if conversations have no winnable state and are not "puzzles". In these cases conversations can be used to flesh out the universe, provide information about it, but have no way to influence the game itself beyond a meta-narrative that you build in your head.

Incidentally this is how most conversations, even with very complex branching paths, work in most games nowadays. Mass Effect pioneered the zero consequence conversation in ME2 after people were very upset that one needed points in persuasion to not let Wrex die in ME1 (a bad design decision for other reasons).

This is an element that, ironically, Deus Ex Human Revolution actually got better than the original Deus Ex. The conversation system is a puzzle, whether you intuit your answers from a personality profile or use the Social Enhancer augmentation, the conversations have a winnable state and are of consequence.

Similarly Alpha Protocol tied decisions and conversation to very specific consequences with benefits and penalties, it followed rules that could be discovered but did not rely on dice-rolls. Simply having a narrative/story reveal is inconsequential, its a nice addition to have on top of game mechanics, but it does nothing to improve the game itself.
A dramatic reveal can give context for your action, and thats a good thing, but if we are talking about game-design that isn't really the primary concern of the tests (challenges, mechanics) you design. Tests are certainly better with context but not limited by it.

I know Mr. Spector is a simulationist but pure simulations usually make for horrible games . Imagine a game where the realistic simulated portrayal of wounds makes the game unplayable because your character needs to recuperate in a hospital for days (MGS3 wounds were breaking the flow). However, simulated elements can be a foundation for great games (see MGS2, Thief), in of themselves simulations aren't compelling gameplay wise (again depending if they can be "played", manipulated by the player)

PS: I am opposed to games being a staging-ground for moral discussion as ethics and morals can not be played. They can certainly feature elements of ambiguity but games, as a medium, are inherently bad at communication, or certainly far worse than more traditional linear media, simply out of the fact that the designer has no monopoly on intent with the systems he makes and that are being interacted with.

Rob Lockhart
profile image
I think there are ways to get closer to simulating a conversation while maintaining a sense of game mechanics and without being delusional about the state of the art in conversational AI. Take a look at games coming out of Michael Mateas' lab at UC Santa Cruz, such as Fašade and Prom Week.

Aleksander Adamkiewicz
profile image
Facade is an example of the delusion.
For a conversation to be meaningful the recipient needs to be aware of what is being said, without it there is no communication.
Facade was a relatively basic algorithm to query for keywords, it had no way to actually understand or even pretend to understand what is being said.

It's like a chat with Cleverbot, the sentences sometimes make sense, but you are fully aware that the opposite is not actually understanding you.

However, I will agree to the point where the game can be "won" by using a consistent pattern of conversation (generally questions), however this test is rather bad, as there is no indication how this can be accomplished or why.

To replicate language is easy, it has syntax, it is formalized. To replicate understanding is not, without it any conversation is empty and its implementation into game mechanics can be easier achieved via keywords and branching paths (old text adventure style).

Conversation systems aren't there yet and wont be in the foreseeable future.

Tomi Vesanen
profile image
I can't help but feel you are completely missing Spector's point in this article. I don't think he is advocating some kind of ultimate simulation experience, but to using a level of simulation that aids in roleplaying - what Spector's after. The point isn't about conversations being like talking to a real person, but talking to a character in the context of a story and a setting. If the aim, like Spector says, is to provide a playground to create an alter ego, perhaps all these puzzles, challenges and attributes are only a hindrance or are not needed to create a compelling roleplaying experience.

Aleksander Adamkiewicz
profile image
"If the aim, like Spector says, is to provide a playground to create an alter ego, perhaps all these puzzles, challenges and attributes are only a hindrance or are not needed to create a compelling roleplaying experience."

The aim of RPGs never was to create an alter ego, thats a relatively new shift/invention (last decade or so). For all intents and purposes "RPG" is a misnomer for stat-based strategy/tactics games.

Historically (DnD, aDnD) the RPG was a mutation of miniature war-games, it changed the context from a competitive game to a cooperative one (vs DM). It also added and formalized non-combat character interaction, providing a more diverse set of approaches to obstacles. Mechanically however there is no difference between Intimidate and Avalanche Strike to cave a guards face in, all that changes is narrative context, but thats separate from the game itself.

If we are talking about games, creating an alter-ego entails these formalized rules in the game-space, once you take away the formalized rules the genre changes to Interactive Fiction, like Non-Linear Visual Novels or Interactive Movies as you are removing the game from the "Role Playing Game".

If Mr. Spector wants a "roleplaying experience" then perhaps he formulated the article wrong, as this experience is not the core of an RPG but only a by-product. Maybe he would find himself more comfortable in the space of improv play-acting like LARP, note that LARP (Live Action Role Play) misses the "G" for Game. Rob for example mentions Facade, which is not an RPG and it is even tenuous if it can be even classified as a game at all.

A game in itself isn't conductive of creating alter-egos. Certainly escapism, numina and immersion can create the experience and lead to play-acting or "roleplaying" but they are byproducts of game-systems or simply provide additional context to systems.

As for Spector not advocating some kind of "ultimate simulation", maybe not in this article, although it seems implied in the Setting section, but anyone that follows him knows that he is a simulationist and argues his game design from that perspective.

Roger Tober
profile image
Conversations are tough and personally, I think keyword is the best way to handle it. That's how we think. I don't really need to go through a list of choices and phony questions, etc. Tell me what you know about ____ little NPC. I'm trying to find it. I don't want to know about your mother unless it involves a "job" you are offering me.

Jakub Majewski
profile image
So... what you're saying is that you have no interest in role-playing games, and nothing to say in relation to the article above? ;)

Seriously, though. I can sorta understand that kind of no-nonsense attitude. I'm kinda in the half-way house in this regard - for example, when playing Skyrim, on the one hand I appreciate the effort the developers put into the conversations, by adding proper player responses. I do think that having to ask questions instead of just selecting keywords adds quite a bit of atmosphere. On the other hand, though, the sad fact is that I click through most conversations at the pace it takes me to read the text. I don't appreciate the voice acting, not because it's bad, but because conversations for me are not there to enhance the experience, but to deliver information.

I think, though, that the main reasons why I treat conversations purely as sources of information is because they are, at the end of the day, done badly - and while not directly for the reasons that Warren Spector talks about, I think the things that went wrong in Skyrim directly flow out of the problems Spector identifies:
- Shoddy characterisation. Many of the player responses just make you wince, because they can be inappriately sarcastic, humorous, and generally out of sync with the world. There is often a big leap between the NPCs' side of the conversation, and the player's side, with the former being polished, adapted to suit the world and the current stage of the story, while the latter is just kind of "yo, I'm the hero".
- Lack of personal adaptation. Choose whom you will, you're still an ignorant fool - that's how the questions you ask are structured. Playing a Dunmer, I have no problems in regards to my questions about Nord culture coming from a position of ignorance. I do have a problem, however, when the conversations force me to plead ignorance of Dunmer customs, or recent events in my character's (presumable) homeland. I also have a problem with my character being ignorant about world-shaking events - when your only option in the conversation is to plead ignorance about the Oblivion Crisis (never mind that the very term is just crazy - the near-end of the world, getting written into history as a "crisis"?), that's a problem. The player does need to be able to ask questions about all these things, because the game cannot presume that the player's and his character's knowledge is the same - but simply rewriting some questions to suggest more knowledge would have made a vast difference.
- Most importantly, lack of significance. Conversations in Skyrim just beg to be treated as insignificant, because it's clear that that's how the developers treated them. Voice delivery is clearly not intended to matter. How many times in real life do we get additional information not from words, but from the way that they are spoken? In real life, this means we must all be careful listeners, attuning ourselves to speech patterns and gestures. In Skyrim, if you read the text, you got all the info you could get, the voice adds nothing (never mind gestures). Another "lack of significance" factor is the simple fact that it never, ever matters what you say. I have yet to encounter a situation where I would have a choice of asking a question in two different ways, and actually getting two different responses. At the extreme end, you can choose between bribing/intimidating/persuading someone... and if successful, they all yield the same result. So, who cares?

Michael Joseph
profile image

Fabian Schneider
profile image
Oddly enough, this has really reminded me of what I am doing wrong in my ongoing Pen and Paper group. Having been a GM for about 15 years, I have noticed a severe shift away from true role-playing towards designing my campaigns and adventures around gameplay considerations. "How do I properly distribute challenges to the competencies present in the group?", "How do I scale encounters to make them fit the power progression I envision in the campaign?", "How do I design a dungeon that is challenging both mechanically and intellectually?".

This is, of course, a flawed way of doing it, and I am fully aware of that. Yet, I can't help but hide behind game mechanics and deceptively easy balancing-by-numbers ("A four-hour session translates to about 4 scenes/encounters"). I imagine that this temptation is even more present for video game designers, since there are technical limitations to take into account that necessitate a certain degree of balancing-by-numbers. At the same time, though, many games have found creative ways around this that make the game feel more responsive to player action.

As far as video game design is concerned, the trick many miss is not to narrowly define a gameplay corridor consisting of only "expected" decisions. In Skyrim, for example - the core focus is on exploring and killing dragons, basically. The addition of ones own house in the Hearthfire expansion was a great idea to provide a path outside the main focus, though it didn't go far enough, thus breaking immersion.

RPGs are about doing what no one expects you to do. By forcing players into preconfigured notions of what they are *supposed* to do (something which strangely happens surprisingly often in sandbox games), one only destroys the role-playing aspect of the game.

Good examples of linear RPGs allowing players to actually play a role are the Baldur's Gate series and Mass Effect.

Ben Mitchell
profile image
While I agree with many of Spector's points here, I think there is a flaw with this approach, and it's a big one. In basically every game that has this type of "yes/no moral dilemma", I don't like any of the options. In fact, I frequently find these divergent story choices actively make it harder for me to role-play. Worse, they hurt my ability to suspend disbelief. And the reason is that they are nearly always false choices; if I were really role-playing, I would choose "none of the above."

One of the big advantages that paper gaming has is that the player can come up with solutions to problems that the GM hasn't anticipated. While this might theoretically be possible in a simulation-game, the systems which are being simulated generally have no bearing on the "moral dilemma" choices (see, for instance, nearly every game made by Bioware, Obsidian, or Bethesda). The standard form of these choices now is actually pretty similar to what Spector was advocating; a "yes/no" (or more generally, "A or B") style of choice. Who lives, and who dies? Do you side with Faction 1 or Faction 2?

Of course, in order to make this a non-trivial choice, Faction 1 and Faction 2 must both have behavior that allows the player to think of them as an enemy; essentially, when each side says "the other guys are bad," they are both right (if one is clearly good and the other clearly evil, it's not a moral dilemma). Every time I see this sort of grey vs. grey morality choice in a game, I just want to scream at them, "You're both wrong!" Who do you want in charge of the world, the faction who will impose a totalitarian police state, or the faction who will allow total lawless anarchy and chaos? Would you rather die by fire, or by ice? Thanks just the same, but neither choice sounds very appealing to me.

There seems to be this assumption that the player will immediately identify with one faction or the other, and then totally ignore how problematic the behavior of that faction actually is. While there are certainly extremists who behave this way in real life, this assumption implies an amazing lack of respect for the audience; do developers really thing their players are this shallow?

There have been a few exceptions to this pattern, but even they are generally problematic. Fallout: New Vegas, for example, let the player side with Faction 1 (NCR) or Faction 2 (Caesar's Legion), but it also let the player reject them both. Unfortunately, the third option didn't actually give the player the option to set up their own faction, it just let them maintain the status quo, which wasn't really any more satisfying; it amounted to simply having three amoral choices instead of just two.

Bioware games have lots of "moral choice" points with two options, most of which are problematic in this way (and possibly none more so than the ending of Mass Effect 3), but they have occasionally gotten it right. The ending of Dragon Age: Origins is a good example; while you can't save *everyone,* you can save *anyone.* No one character is fated to die, but some sacrifice has to be made. The narrative reasoning for the sacrifice is clear and strong, and telegraphed well ahead of time; the choice does not come as a surprise to the player. Because it's the end of the game, the player can choose to sacrifice herself, if none of the other sacrifices are deemed acceptable. The NPCs are all willing to sacrifice themselves; everyone goes in to the situation with their eyes open, and the NPCs volunteer for the sacrifice, rather than being forced or coerced (either by the player or by the situation). Best of all, there is an explicit requirement that a sacrifice must be made to achieve the goal; no one dies because the player wasn't fast/strong/smart enough (trying to make the player feel responsible for an unavoidable event which was scripted by the developers is both cruel and insulting). Also, the mechanical rules of the game-world are not violated or ignored in the process (this is frequently the case in settings like DnD where a spell like "Resurrection" is commonly used when a character dies during the course of normal gameplay, but is unavailable when the same character dies as a plot-point; e.g. the scripted death of a party NPC in Neverwinter Nights 2, the original ending of Fallout 3, and pretty much every time a character dies during a cutscene in any JRPG). Sadly, Dragon Age 2 is a brilliant example of what not to do if you want to have satisfying moral dilemmas, as it has nearly all of the problems discussed above, and on top of them nothing you do ultimately matters at all.

Eric Robertson
profile image
I agree with the need for smaller areas that encourage more social interaction.

I believe the players are content for other players. This includes what they say, what they build, what they destroy, and how they behave. I believe we as developers can encourage this via rewards for 'any' interaction and hope the players don't eat each other, or at least not too often.

Erin OConnor
profile image
This is exactly why I love FFG (Fantasy Flight Games) Star Wars Edge of the Empire so much.

The [sadly] no longer available Marvel Heroic Roleplaying by Margaret Weis Productions was pretty decent too.

(Damn, this sounds like an advertisement)

Jean Louis
profile image
Hi Erin, any experience with Saga Edition? If so, any thoughts on the differences in how the issues mentioned in this article are handled better/worse by Edge of the Empire? Thanks!