|
[Can games be realistic emotional simulators? Game designer and writer Ernest Adams explores in depth how AI might actually emulate human emotions.]
Computers were first invented to crunch numbers for their
own sake. Charles Babbage's Difference Engine was built to calculate
polynomials, and Lady Lovelace wrote the first known program for his Analytical
Engine to compute a sequence called the Bernoulli Numbers.
It wasn't until later that computers began to be used for
applied tasks such as figuring out trajectories for artillery shells. Later
still, computer games were invented. We put a veneer of meaning on top of a
mathematical simulation to create an imaginary world, but often the veneer isn't
very thick. It's particularly noticeable in economic simulations and
role-playing games. You win the game by watching the numbers on the screen.
This makes sense in economic simulations, especially ones
that are supposed to serve an educational purpose. Recently, I got a chance to
play a beta version of a funky little game called Super Energy Apocalypse from Brain Juice Games. Its designer, Lars
Doucet, characterizes it as "Sustainable energy use... AND ZOMBIES!"
Every night your community gets attacked by zombies
(actually, garbage-eating aliens disguised as zombies... it's a long story). To
win the game you have to develop non-polluting energy sources to power your
defenses. Zombies love smog and nuclear waste, so you have to keep your eye on
both the energy production rate and the amount of smog and waste you're creating
in the process. It was fun, and, as it's based on real-world power generation
systems, I learned some things.
For the most part, though, I'm starting to find myself a bit
tired of games in which the mathematics are so close to the surface. I love
role-playing games for the exploration, the stories, and their large variety of
characters and locations, but I'm less enthused about the constant trading and
upgrading.
All that emphasis on gear seems distinctly nerdy to me. There's not
much difference between bragging about your Superior Glowing Voulge of Major
Whacking, and bragging about your overclocked liquid-cooled Alienware PC with
the Go-Faster LEDs in the case.
RPGs are about character growth, but it's all character
growth as expressed in numerical terms. At the end of the game your character
is faster, stronger, more dexterous, and so on; you have the figures to prove
it. But he might as well be a robot as a human being. There's precious little
of what we might call psychological character growth -- growth as a person.
So I'm starting to think about games that turn numbers into
behavior. Of course, many games use finite state machines for their characters'
AI, and the finite state machines take numerical values as input. ("If the
enemy is within range, switch to attack mode.")
But these machines don't
change -- the characters don't grow and learn to change their behavior. And as
designer/professor Michael Mateas pointed out, finite state machines can't walk
and chew gum at the same time -- i.e., they only exist in one state at a time,
so they can't exhibit complicated mixtures of actions.
 An early example of a "character" apparently
changing its mood and behavior in a sophisticated way appeared in Chris
Crawford's Balance of Power. I've
written often about this game because it was both brilliant and unique -- a
simulation of superpower geopolitical maneuvering that, so far as I know, has
never been tried since.
It was a single-player game. You played either the USA
or the USSR,
and you tried to maximize your country's prestige at the expense of the other
side, in part by facing them down in a series of diplomatic crises. (The Cuban
Missile Crisis is the most famous real-world example.)
In each turn, both sides undertook various diplomatic
activities (such as forming alliances) in smaller countries around the world,
and in the next turn, each side had the chance to challenge the actions that
the other side took -- a crisis. When a crisis erupted, one side or the other
would eventually have to back down. It was essential to choose your battles
carefully, because if you provoked the other side into nuclear war, you lost
the game.
Balance of Power was
such a hit that Crawford wrote a book explaining how it worked, right down to
the equations he used. The book was called Balance of Power: International
Politics as the Ultimate Global Game. It's long out of print, but Crawford has
made a plain-text
version available on his own website. I think it's a must-read for any game
designer: a clear and intelligent explanation of how mathematical formulae turn
into behavior.
|
Printer-Friendly Version
Yup, this is the challenge, however it can be balanced with elegant and clever design. Scripting is too asymmetric. To borrow a phrase from Craig Perko, it´s like trying to like a sofa with a toothpick.
Anthony said that humanized NPCs would alienate most players, but I don't think that's a valid stance to take. There's no proof, for one thing. The same types of arguments have been made at offering different types of content in other media such as film, or even licensing products such as Japanese entertainment targeted for female audiences ("oh, girls and women aren't interested in such stuff in the English market" was a very common claim 10-15 years ago). It may be that certain groups of current players, perhaps even a majority, would reject such content, but that's irrelevant because offering such content might appeal to people who refuse to play right now due to the lack of humanized NPCs. This idea is what has propelled Nintendo's success with the Wii and DS; don't compete for market share out of the same barrel with Microsoft and Sony, just go get market share from a different barrel.
I've often argued that the much more humanized characters in many Japanese adventures, visual novels, and simulations are precisely why these games have greater appeal to me, personally, than much of what is created in the English market. The same thing applies to English comics and animation compared to their Japanese counterparts. This appleal isn't for everyone, of course, but the growth in popularity of such entertainment around the world over the past 10-15 years has demonstrated that there was certainly an untapped market demand. I think that the arguments in this piece are very similar in nature; offer content that appeals to people who find current offerings to be uninteresting or unappealing.
Quick point about The Sims, of which you state "The player doesn't get any convenient digits or power bars. She's not supposed to be thinking about the game's characters as simulations, but as real people." Not quite true, since The Sims 2 provides a wealth of quantification of your Sims' needs and so on. In fact, in my doctoral research it was pretty typical of male players of the game to be more or less obsessed with playing The Sims like an RPG, maxing out the "stats"...
More generally, the only thing I'd like to say about this debate is the importance of remembering that we don't need to *really* simulate emotions and so on inside the algorithms of the game, we need to do enough to help players *interpret* what they see as involving complex emotions and so forth. I'm guess that's what Balance of Power (I've not played it, sadly) does well. I also think that people are only too willing to attribute complex psychologies to things they don't understand the inner workings of... so in theory that's win-win...
Tons of games, even in the mainstream, are HUD-less. That's one definite step closer to immersion and avoiding stat stacking for the sake of it (if stats became invisible, like the HUDs).
But character development could be reflected in a more linear fashion first. For example in Zelda, Link never emotes anything when enemies approach. As a sign of an enlightened Link a simple battle ready stance that wasn't as confident at the game's start, could show development.
Wind Waker had some signs of this through struggles pushing big boulders or determination when swinging swords.
My interpretation of the article was that the author was advocating a more complex and multi-faceted relationship between player and NPC, where the effect of player action on NPC was not readily apparent. My take on that was that a game as described would be maddening to most traditional gamers because the correlation between cause and effect and the reasons for success and failure would be difficult to determine.
I feel like you read the article to be a call for more realistic characters in games, which it is not. I even addressed this when i said, "video game characters should exhibit more human behaviour, but as written by the developers not as calculated by formula."
Perhaps there is a segment of the market that would find a human simulator type game interesting. From what I understand of D&D the rules can be kind've amorphous, so maybe theres a model for success.
The Sims do have stats, but the relationship stats are only shown in a crude manner (it shows what your character feels about the other person, not how the feeling is returned). Because of this, even if the relationships of The Sims are superficial and simple (it's only love or hate, and mostly based on money, stats and power and don't consider the other relationships) you mostly handle the relationship in a way you'd do so with a person, you expect similiar reactions even if you know it's nothing like the gamut of feelings.
D&D on which most if not all RPGs built (hence the leveling and stat system) doesn't have any system for handling people, it lets the DM handle this based on his human experience, computer's don't have that, so a crude system is invented, this systems are ridiculous because by being so predictable they give unexpected inhuman results. Example: in Fable you could be as evil as the Devil, pay a great amount of money to the church, become goodly and nice and everyone would believe you were good. You can do this as much as you want without anyone suspecting anything. This is not what you'd expect from people, so you stop treating them as people, and since the story depends on emotional tying to the characters, little if any occurs. What if some characters mistruted you? What if some characters had mixed feelings because even when you are good you do certain neutral actions that they feel are evil? You don't have to have full emotions and realistic simulations of feeling, but enough to trick the player into thinking that the feelings are real, and to treat those NPCs like characters, not graphics on top of a bundle of numbers.
Let me give a more concrete example. In the football game I worked on, we would transition the player from the end of play back to the huddle. During this time, the presentation engine had to choose between field shots, player/coach reactions, audio commentary, and so forth. In many instances, it was this unseen "director" that felt lifeless and robotic. Either because the focus was completely wrong (focusing on something relatively unimportant while skipping a minor but emotionally important moment with a given player), or because we'd go to a cutscene with much higher emotional level than was actually warranted at the time. How many times have you seen a game character who clearly had three too many espressos for lunch?
While many games have an exceptionally limited presentation (attach camera leash to player), others are ripe for significantly richer presentation. I believe that emotional intelligence (and the understanding of primitives like storyline, history/memory, wisdom) would make for much more compelling games.
Breaking the Cookie-Cutter: Modeling Individual Personality, Mood, and Emotion in Characters given by myself, Phil Carlisle, and Richard Evans (coincidentally of "Sims 3" fame).
We will be discussing exactly the points raised here including some of the pros and cons. We will discuss some examples of how infusing your characters with personality, mood and emotion can be accomplished technically, artistically, and offer deeper, more immersive gameplay to your players.
On topic, it is possible to craft these issues into your games without "maddening" your players. The trick is, if the reactions make sense to us as people, they make sense for our characters as people. If done wrong (or arbitrarily), THAT is when it alienates people. If done right, however, it can be an extraordinarily powerful tool for drawing people in.
And after all, how many times do we have to read reviews saying "the characters felt dull and lifeless" before we stop parroting the mantra "our players want predictability"?
You raise a valid concern, and I'm sure some players would indeed be put off by more complex (and therefore less predictable) behavior from NPCs. I would argue however that it certainly won't put off everyone; many people enjoy multiplayer games where the other players can be incredibly unpredictable, so there's no reason more complex NPC behavior couldn't also be enjoyable. Personally I'm all for trying it out to see how real players react.
It seems to me that this level of AI cannot be arbitrarily hammered into existing game designs. The additional complexity introduced by it needs to be appropriate to the role the AI plays in gameplay. So, even if we do have the technology ("better, smarter, more human?") we need to identify the complexities added by each new facet of behavior, and the impact on core gameplay.
Of course, the other side of this is that these additional complexities give us opportunites for changing our gameplay for the better, as long as we are aware of them and design to use them rather than just trying to make them fit what we currently have.
A recent example of this is Left 4 Dead, which uses static levels but a variable items and enemies placement system (which is apparently connected to how well the player is doing using some fuzzymaths) and it is pretty effective for a while, keeping you in the scary zombie movie chic far longer than most games of that type too.
On the other hand, there's a point at which such subtlety becomes pointless. One of the criticisms oft noted about Facade is that it feels quite abstract. There are lots of things going on under the hood, but they never really translate across to the player. Where games lose their point of playing (clear goals) cross over into what amounts to pure simulation, they become very reliant on player-driven goals, but player-driven goals are generally just not that fun.
That's when things become nothing but a sandbox and the average player interest drops massively. I grant that there are some players who really do enjoy the creativity of a detailed simulation just to see what they can do with it, but way more players generally can't be bothered putting in the effort to learn the skills to get to the other side of entertainment.
At this point people will often bring up games like Spore or The Sims to dispute this, but in my view they are overstating the impact of both. Many are the stories of players that trudged through the first sections of Spore just to get to the space sim, found it rock hard and stopped playing. Many are the theories of why The Sims is so brilliant when 95% of that game's appeal is essentially RPG mechanics cloaked in a veneer of suburbia.
It's a great exercise for developers to play with the idea of complexity and the scope for complex results but it often doesn't really translate to players, and thus is generally bad design rather than good. It's hard to judge when subtlety' point of diminishing returns is too approaching, but it tends to be nearer than we would like.
But I thought the answer to this would be more scripting, not more simulation. Taking the math away from the player actually would help obscure the switch between simulation and scripted events, allowing designers to script material on top of their simulated "character arc". That is the opposite of Façade.
Don't get me wrong, I think AI storytelling is a good goal for experimentation, but look at the example on the article itself about the hidden variables controlling aggression on Balance of Power. Not displaying them (or telling players they were there) allowed the game to feel it was acting in an emotional way because they didn't know their own actions were influencing the outcome.
Imagine the Karma and Luck systems in Fallout 3 were not exposed to players. The scripted events in the game's story would feel like proper retribution and, what's more important, players wouldn't feel they are being punished for their actions by restricting conversations due to low or high Karma. Obscuring the statistic, the options the player is not experienced *never actually happen*. They couldn't. The reality of the game didn't include them.
That means even if the episode that has been triggered in this manner is a pre-set script selected among others players will have a harder time being able to determine that. The designer's job made easier, not harder, by incorporating these techniques.
By simply having an AI character make their decisions based off attributes, and those decisions changing those attributes, you should be able to achieve the mutable situation the article describes.
In any case, it's heartening to know that there are other people out there thinking along the same lines - it means I may be going in the right direction!
To address the fear of "angering (traditional) players", several points need to be made. What the author proposes is to add another tool to our Game Design toolkit. Like all the tools in this toolkit, there are situations where they can be used and those where they can be misused. Consider these options:
* On which AI are these emotional responses implemented in the game? If the AI who have these emotional responses are ambient, like the people wandering the streets in GTA, then I welcome interesting, emergent emotional behaviors. Sure, I don't want the target of my "sniper mission" to have a Freudian breakdown that fubars his pathfinding. However, I would consider that a bug to be fixed, and not a reason to abandon this approach entirely.
* What type of game are you applying this tool to? Judging this approach when applied to Indigo Prophecy is not the same as judging it when applied to Enemy Territory: Quake Wars.
* Anthony's concern takes it as a given that a system like this would have sole control over AI for the duration of the game. In fact, there's no reason why designers couldn't switch between this procedural method of AI control and a completely scripted version when it suits the story and gameplay.
Although there are limited applications of this approach when you attempt to shoe-horn them into existing games, it doesn't take a psychologist to see how future games could give players what they've been asking for year after year - an experience that's not a rehash of last year's hit.
i haven't read every ones reply so i dont know if what i'm about to mention has been mentioned or not. i've thought for a long time a system that would fit the emotional model of video game development well would be to develop a "conversation engine." it would be a chat dialog simliar to SCUMM games, but would construed to encourage fast game play and dynamic conversations - not just static dialog trees. These dialog trees could be used to examine a given NPC's state in the machine state system, and thus alter their state through dialog. A game could be construed around this setup, and in theory could create a more dynamic but socially and emotionally centered game.
One thing they discovered, is that players could not read the emotional states of their characters if they tried to present too many aspects of that emotional state at once. They ended up using presentation techniques that suppressed the contribution of all but the currently-strongest one or two emotions. So even though a complex mix of things was being induced in the characters, one emotion would win out and be presented clearly to the player (via stance, facial expressions, gestures, and so on).
Anyway, their publications are still available here:
http://www.cs.cmu.edu/afs/cs.cmu.edu/project/oz/web/papers.html
A more recent article here on the AI difficulty in Pure is also a good example of how emotion can help with gameplay. While the article is about difficult and not emotion persay, it does offer thought on how one can model hidden behavior into something relevant to the player. One may not need to have numbers for likes and dislike so much as model the appearance of NPCs wanting and doing along side the players doing the same.