What’s your job at Bethesda?

[elander_]

Not when the tool fails to do the job as it was advertised. Mind you I think we're happily criticising both.

In what way did RAI as a tool failed to do what it was supposed to do? If you are talking about path-finding i never saw any npc having any problem with it in my computer. Everything else can be tracked to bad AI setup or a lousy sense of realism from the part of game designers.

 

[Mayday]

Perhaps you didn't play long enough.
I remember monsters blocking on low boulders, allowing me to constantly take them down with a bow, because they couldn't approach me.

And how isn't bad AI setup a part of the AI system again?

"In what way did RAI as a tool failed to do what it was supposed to do" DU has posted the answer to your question before you even asked it.

 

[Naked Ninja]

How were the schedules good? I've only seen people do these things: walk around, sleep, talk to each other about rumours and eat a food item pulled out from the air when at an inn (they weren't even buying it). I've also seen ONE person read a book- so that was most probably hard coded.

You didn't see guys go work in the fields, guards patrolling, thieves robbing merchants, guards protecting characters who are being robbed, mages practicing magic, people going to church and praying, guys going to the park, nocturnal characters whose schedule was based around night time, People going home after their days "tasks" to eat then sleep, practicing archery,shop owners following you around suspiciously when you went into their back rooms?

Slightly more than just wander->sleep->eat->rumor.

 

[DarkUnderlord]

... the potential is still there. The lameness is that they didn't pull it off like they said they would, but that doesn't mean that they couldn't have given more time.

• Quests are the bedrock of Oblivion's game mechanics. You can't really play the game without encountering quests. Even if you decide to be a rebel and ignore every headlined quest, you will still encounter them.
  
  For example, let's say you decide to use your character as a hunter of beasts and a gatherer of plants. You then sell these to Merchants.
  
  When you encounter a merchant and enter the barter/sell menu you are in a quest. The game has a quest, which does not appear on your journal, which controls merchant transactions. When you walk down the street and overhear two NPCs talking, the conversation is controlled by another hidden quest.

Sounds like every other game to me.

• The movement AI in Oblivion is assisted by a set of path grids. Normally the render window does not display the path grids. To see them, you can select the Path Grid Editing icon from the tool bar, or you can select Edit Cell Grid Path from the World menu.
  
  For normal movement, actors always use the path grids. The game evaluates the path grid to find the shortest path, and sends the actor along that path.
  
  If an actor is told to go to a specific spot, usually indicated by a marker, he will go to the nearest path grid point, and then move straight toward the marker. In combat, if an actor has direct line of sight, he may choose to go straight toward his target, ignoring the path grids.

Sounds like every other game to me (though I believe Half-Life called them nodes).

I think the main point about the RAI system is that it's an efficient way to handle scripting. Create packages that add functionality to the NPCs and then it's as easy as assigning them.

"Package" is kind of like Oblivion's word for "a bit of code". I can (quite literally) do the same thing in Arcanum. Setup a script which calls a "package of code" which handles how the NPC gets from point A to point B. All I have to do in my script is set a destination and give it a condition under which it triggers, the AI "package" handles the rest.

• Packages are bundles of AI instructions with conditions for when and how to execute them. They are placed on an actor by dragging them to the AI Package List. When an actor needs to pick a new package, the list is examined from the top. The first package that is valid based on it's schedule and conditions is selected and applied to the actor. Exmaples for packages you may have witnessed in the game are Sleep (find a bed and sleep), Accompany (e.g. an NPC following the Player), or Eat (find food and eat it).

This is where you might be thinking "wow, the AI doesn't need to be told what bed to sleep in, it'll just find a bed and sleep". And it is pretty wow... Until you find out that in Oblivion, most of the NPCs had their home beds tagged (otherwise, it'd be pretty stupid for an NPC to sleep randomly in some other bed) and in Arcanum, you do the same thing. You tag a bed for a given NPC and come night time, the NPCs wander to their tagged beds and sleep. All you have to do is tag the bed. Why, that's simply amazing.

If anything it was a failure of Pete Hines to properly portray it for what it was in terms of the game-play. It was, however, capable of doing what they'd said. There just wasn't enough time to refine it enough.

I don't think you really understand how Oblivion was sold or how the AI actually does what it does. "NPCs will steal food if they're hungry!" something which I believe was reduced because they, like every other game before them, couldn't figure out how to deal with the consequences of that without making it look stupid. Unlike every other game before them though, they decided to implement their half-assed version, rather than holding out for an "all or nothing" approach.

Truth is, Oblivion's "Radiant AI" handles things the same way every AI in every other PC game does before it. Certain blocks of code are setup as functions. You pass certain variables to those functions via a script (EG, the ID of a spell to cast or an action to perform), you setup various objects by tagging them, the function handles the rest. The only real difference is that in Oblivion, the programmers took the code one step further and instead of just having "find player, kill player", they added in "find food, eat food" (though I'm pretty sure most of the food was tagged as well, given most NPCs wander to the same bar they always do and eat the same things they appear to have eaten the day before. I haven't poked around that far into it to find out though. Mind you, they do appear to produce mugs of ale out of thin air when required. I might be wrong about that though).

In what way did RAI as a tool failed to do what it was supposed to do?

http://www.gamechronicles.com/qa/elders ... livion.htm

Oblivion features a groundbreaking new AI system, called Radiant AI, which gives non-player characters (NPCs) the ability to make their own choices based on the world around them. They'll decide where to eat or who to talk to and what they'll say. They'll sleep, go to church, and even steal items, all based on their individual characteristics

Most NPCs talk to anyone they bump into (I think it's a faction thing). They "decide" where to eat by being told where the bar is. They "sleep" because their schedule says "sleep" and they sleep at home because they're told where home is and they're told to go home and sleep there. They go to church because someone put it in their schedule and told them where the church is. They steal items because they're told to. Somebody correct me if I'm wrong.

http://planetelderscrolls.gamespy.com/V ... tail&id=23

In regards to the new Radiant AI system, it has been stated that NPCs will be able to think and react independently of scripts. Does this mean that a player could order an NPC to do something (if in the position of a guild head, etc.), or perhaps find a random unscripted quest due to independent NPC actions?

Yes, we can do those things. I''m not saying they are in there, and we're toying now with watching NPCs do things and how we can really get the player to affect that or have more fun with it, or even see it. So I won't give specific examples right now, but we'll be trying some similar things in places. I can tell you that our goal for the Radiant AI was the “Fargoth” quest in the beginning of Morrowind, which took some heavy scripting to get Fargoth to behave well, sneak around, steal the ring, put it in the stump, and such.
Our early goal for the Radiant AI was that kind of thing just “happening”, without any scripting. And it works - which is great. But if we didn't tell you what Fargoth was up to, you would have never noticed, or it would have looked really odd. Anyway, that's the stage we're at, we have the behaviors, and we're trying to maximize the player's perception of what's happening.

"We built a system that does these amazing things (trust us, we're being honest!) but we decided not to use them." There are really only two conclusions you can draw from that, the first being they built an amazing system which they couldn't get to work in a sensible way, so they scaled it back so much that it has little affect on the game; the second being it is in there but it doesn't do any of the amazing things it's meant to.

Whichever way you look at it, there is no "Radiant AI" in Oblivion at all. It's either there and they didn't use it or it doesn't work as advertised.

 

[elander_]

Perhaps you didn't play long enough.
I remember monsters blocking on low boulders, allowing me to constantly take them down with a bow, because they couldn't approach me.

That never happened to me. That was my experience with the game. Maybe you had a different experience because you have different hardware.

And how isn't bad AI setup a part of the AI system again?
"In what way did RAI as a tool failed to do what it was supposed to do" DU has posted the answer to your question before you even asked it.

In the context i was discussing it's not the same thing. I was discussing RAI as a tool. I could reproduce everything you mentioned from Ultima7 easily in Oblivion with a mod.

Whichever way you look at it, there is no "Radiant AI" in Oblivion at all. It's either there and they didn't use it or it doesn't work as advertised.

DU i can agree with that. It doesn't matter if RAI is a nice tool. In the end, for the player, it's worst than something like Ultima7 or Gothic. I just don't see the point of blaming the programmer for the screw ups others did with the tools she built. If there's someone to blame is Todd and company and his weird sense of reality.

 

[DraQ]

At very best the RAI in Ob is good, but grotesquely misimplemented AI system. At worst it is the very crap it looks like.

What's so revolutionary about RAI anyway? That it checks for some context dependant variables before applying the scripts?
It still applies it's scripts as mindlessly as Doom did, staying well behind some pre 2000 games in situational awareness department. It still needs pathgrids for pathfinding. It doesn't recognize audio cues which was successfully implemented in much older games. It still doesn't react to corpses. It still walks blindly into many environmental hazards. It still doesn't know how to jump. It's still completely incapable of any tactics beyond "hit it with projectile" , "run away" and "run up to it and hit it with blunt/blade" and woefuly predictable in it's actions. It's liable to hilarious fuck-ups, like sneaking up to someone in broad sunlight crouching right before the victim and trying to pickpocket them. It's far surpassed by almost 10 year old Unreal in cunning, by over 10 year old TNP in spontaniety and by 9 (I think) year old Deus Ex in situatiounal awareness.

Who cares if it uses somewhat more object-llike appoach to AI scripts?

You're being a trendy douche bag if you criticize someone for the way they look. Post your own picture and I'm sure many of us would laugh too. That's not the point though, the real issue is that the Codex doesn't need morons of this nature. Post something intelligent or get the fuck out.

QFT.

 

[Helton]

I retract my heartless implication from earlier. She's a beautiful flower and I am a shallow trendster. I hope my fellow trendites will join me in recognizing our horrible fault.

 

[Section8]

I have a problem with that statement Section. You're assuming you CAN sit down and map them all. I have programmed AI systems, and my experience is that because of their limitations it becomes a choice of how much independence you want to sacrifice to ensure the system is stable and predictable. Limit it a lot and you ensure it won't freak out, but you limit it's ability to adapt (the scripting end of the system). Give it more free reign and you increase it's capabilities but also increase the chance of it going fruity.

Oh, I don't suffer any delusions of trying to map out every possible permutation of a dynamic system, but...

If you give your AI the power to be "offended" by an action and react accordingly, it doesn't take a genius to figure out that lethal hostility can't be the only valid reaction - even if only because you have invincible agents in your system, or just because you don't have any systems to increase population.

In the end, I feel Bethesda left themselves no choice but to reign it all in and have a system that is mostly script driven, just because they didn't really consider ANY of the probable results. So we ended up with a bandaid fix to something that was always going to be horribly broken - very much like the pause "feature" of Baldur's Gate.

 

[Naked Ninja]

If you give your AI the power to be "offended" by an action and react accordingly, it doesn't take a genius to figure out that lethal hostility can't be the only valid reaction - even if only because you have invincible agents in your system, or just because you don't have any systems to increase population.

Actually, you are mistaken, they did implement that. That is what you are seeing in this :

http://youtube.com/watch?v=MVAYY5LHhv4

The NPC steals a loaf of bread, the surrounding AI realise and triggers a message reacting negatively. However, they don't take any lethal action. Then later the AI is in random comment mode and starts up some gossip with the same character. It comes across as very disjointed, doesn't it? Because they are using AI to simulate human interaction, which is problematic because human interaction has a vast array of subtleties. In real life the dude might have been shoved, or spat on, or told to leave the bar or just glared at. The AI doesn't have that and so comes as horribly disjointed. This is why I laughed about the uncanny valley comment. Oblivion isn't even close to uncanny valley.

In fairness adding that kind of content dramatically increases both content development costs and the cost of testing. And because it won't even come close to human intellect you will still get funny shit happening. The problem is it takes incredible complexity to simulate human social systems and we don't have that. You aren't going to be able to program an AI that a human observer won't notice is fake pretty quickly.

BTW, I remember in Gothic 3 NPCs reacted pretty pretty harshly to theft too. I generally got killed trying. Didn't seem a particularly superior solution. Oh, and the repetitive dialogue? If i count how many times I heard "It's the same old story" in Gothic 3 it easily equals the Mudcrab lines in Oblivion. It was just less intrusive seeming because it is less memorable a line. Which is probably what Beth should have done but they were AIMING at more realistic human interactions and conversations. They failed, but like I said, I think they are a company that at least tries, even if the goal eludes their grasp.

 

[Slylandro]

@Naked Ninja, that's just a plain case of sloppy programming, it has nothing to do with how complicated it is to model humans. It would've been pretty easy to add some state to each NPC to make sure that it didn't randomly gossip with other NPCs it saw thieving/killing/<insert> earlier and instead reacted with <insert> depending on its personality (if Oblivion NPCs have actual personalities which I don't think most of them do). And the thieving NPC should avoid coming near those who saw it thieving earlier. Now, this isn't exactly a perfect model of human behavior. The thieving NPC could come near the witnesses later to try to murder them, while they're in bed and no one is around, for example. This is complicated behavior, and misprogramming this could be attributed to the complexity of the task. But not something simple like two formerly antagonistic NPCs chatting it up about the walls they recently saw.

 

[Nedrah]

Gothics way of handling player theft... well, I actually do believe it is superior. You are going to make DAMN well sure that you will not be caught. If you do get caught, lethal force is applied, which fits the whole brutal, anything goes medieval setting that it does pretty nicely. If you just walk into someones hose in G2, them and their neighbours will approach you with blades in hand and tell you to get the fuck out. If you don't, they stab you in the face. How is that not superior?

Of course, it does have its weaknesses. I remember first entering the harbor district in G2, only to get bullied by some asshole. So, I told him to go suck a dick and started running, because I knew I had passed a guard less than a minute ago.... who was still there, yes, but did for some reason not have any reaction for me talking to him while being chased by some criminal scumbag, just his standard lines. One of the few times I got pulled out of the experience pretty harshly. Things like that and much worse ones where you couldn't make up an exclamation even if you wanted somehow happen every 5 minutes in Oblibion, though - and get hillariously bad if and when you ever decide just to follow some npc around for a while.

 

[Lingwe]

Having modded for Oblivion I have to say that the Radiant AI system is reasonably simple to implement. Once you figure out what all of the different packages do then you can make NPCs with quite detailed schedules.

The main problem with Oblivion that I see is that the amount of work prevented the designers from giving every NPC a detailed schedule. That is why you will see so many NPCs who have nothing but an eat, sleep and wander package.

While it may just be scripting organised into packages it makes it a lot easier for modders to understand how it works.

 

[Darkflame]

The Witcher AI is miles beyond Oblivions, and AI wasn't even that game's main selling point (unlike Oblivion). Just shows you how much Beth sucks.

Witcher NPCs:
have conversations (better than the stupid Oblivion ones) / eat and drink / play instruments / rake/chop wood/ forge weapons/ build houses/ take pisses/ duel with eachother/ play tag/ run out of the rain / pray / light fires / play dice / and probably a bunch of stuff I haven't seen yet.

Oblivion NPCs:
have stupid conversations over and over again / run into walls / get confused and fight eachother / stand in the rain / don't do half of what Witcher NPCs do.

In fact, the only unique action Oblivion NPCs do is they loot corpses. however all this entails is them bending over the corpse - anything they take is added to the game via script, while the corpses inventory remains fully intact; so in other words they "loot" corpses but don't take anything. So they really don't "loot" corpses either...

 

[Jasede]

Sounds like the Gothic "AI" tbh. [Witcher]

 

[Raapys]

Hmm, the Raident AI system is not really an 'AI' by itself, it's simply a way of making scripting common NPC actions faster, as Lingwe says.

The quality of the AI still comes down to the work put into the NPC by the creator, and this is where, unsurprisingly, Bethesda have failed. In the Gothic games each NPC has been put far more work into than the shallow NPCs in Oblivion, and it shows.

Then again, it is probably a somewhat unfair comparison, since Oblivion does, after all, have about 1000 NPCs, whereas Gothic 1 and 2 has...well, I don't know, but I'm guessing no more than 200 for G1 and perhaps 400 for G2. On the other hand, Bethesda have way more manpower. *shrug*

 

[Xi]

...

Yeah, I agree with what you're saying I just think it wasn't as terrible as people make it out to be. Certainly not as grand as Pete Hines would have us believe either. I think the real issue is that AI isn't that interesting in action. It's artificial and is an attempt to mimic a smarter human mind. It's so chalk full of pattern that we see right through it no matter how complex it is because the truth is that it's very simplistic on the inside even if it appears complex at first glance.

Oblivion's RAI is no different, it's just a good way to continue development of AI in a meaningful progressive way rather than reinventing with each new game. That's its real potential. I'm sure there are other developers who've used a similar approach too. Probably nothing new! Still, for a junior programmer who's not specialized in AI development, it's not too bad really. That was the point of my post. It's more than most of us could hope to accomplish given the task. /shrug

Edit: AI is just a predefined pattern no matter how you look at it. Even if there is a dynamic element, the different branching outcomes are very limited. It's just not that spectacular in the end. It's easy to say that this game or that game has "better" AI but deep down that's just not true. It just appears a little bit more complex but is truly simplistic. For a generic AI system RAI does its job well. Most of the other games people are siting for examples of "good" AI are merely very defined scripted events where more care has been given to a specific NPC. RAI wasn't trying to be that specific with OB and its potential lies in the fact that eventually its generic packages will become nearly as complex as direct scripted events with its generic assignments. That would be an accomplishment and that's the progressive element of RAI that makes it good. It can continue to be built upon unlike other games.

... Eventually a system like this could define the difficulty of an encounter where the more packages that are added the more difficult the encounter becomes. Hell, you could create a shit load of predefined patterns that interact with one another in complex ways that create AI challenge that out dates simple statistical stat/level challenge. That is what people are interested in. Leveled loot and leveled enemies is pretty pathetic in terms of creating challenge. It seems like it would be better to create complex AI that is assigned to do this. Only real problem is that this shifts the character-to-player skill ratio of an RPG more onto the player where it's the player overcoming an outcome instead of the character. That's problematic in terms of an RPG, but it's still interesting and RPGs can adapt to this I believe. It may also enhance the "choice" game-play that RPGs are all about where player choices become more meaningful as every choice has direct consequence in terms of dealing with an intelligent(or what appears intelligent) encounter.

 

[kris]

If you want to make a believeable world and believeable characters then you should best use the system that works best in giving that. Illusion or not. for aplayer the intelligence in the actually system never can compete with its actual performance.

Players should call it as they see it.

Yeah? I found the generated universe of Space Rangers 2 less engaging than a hand scripted setting would have been. LET ME NOW RIP OFF THE GUYS WHO PROGRAMMED THE SR2 AI, HAHAHAHAHA, I'M GOING TO POST UP THEIR PICTURES AND MOCK HOW THEY LOOK, HAHAHA.

Maybe you quoted the wrong guy here. Your answer to my post seems quite strange considering I never played space rangers 2. I can only tell that my brief experience with Oblivion gave me a impression that the AI didn't really do anything impressive at all. Mostly I saw people seemingly walk around with no purpose, like everyone in the city where cruising people without work. At least the farmer was working the field though, almost the only character I actually could tell who/what he was.

even in Fable things seemed better as people looked like they had a role in the world. It may be helped by the overall impression of how things where presented, something I think Oblivion failed terrible in.

 

[YourConscience]

Edit: AI is just a predefined pattern no matter how you look at it. Even if there is a dynamic element, the different branching outcomes are very limited. It's just not that spectacular in the end. It's easy to say that this game or that game has "better" AI but deep down that's just not true. It just appears a little bit more complex but is truly simplistic. For a generic AI system RAI does its job well. Most of the other games people are siting for examples of "good" AI are merely very defined scripted events where more care has been given to a specific NPC. RAI wasn't trying to be that specific with OB and its potential lies in the fact that eventually its generic packages will become nearly as complex as direct scripted events with its generic assignments. That would be an accomplishment and that's the progressive element of RAI that makes it good. It can continue to be built upon unlike other games.

While I agree with you on the other points, I really cannot on this point, an AI just being a predefined pattern. It's just that most game developers tend to add AI as an afterthought and yes, making it just a predefined pattern seems to be the easiest and fastest and most straightforward way to do so.

However, some games show that even a patterned AI can produce very life-like simulations - imagine playing one of the heroes breaking into a dungeon keeper dungeon (or evil genius). All those critters busily mill about and do things that make sense and you could disrupt the economy of the dungeon by stealthily moving about and just killing the puny workers. In fact, a game like this would be cool. And that's still just pattern AI.

Then there's more interesting, developing AI, for example in that old game called 'Creatures'. Essentially, in a more generelized way you don't have to give an AI patterns. You only have to give it a model of goals and success/nonsuccess evaluators as well as some kind of counting mechanism which would count the co-occurrence of certain action chains with success and nonsuccess in the end.Depending on the complexity of the gameworld running a simulation would in the end produce the most efficient behaviour of you AI agents, given the game rules.

But, whereas this might sound nice and cool, setting it all up such that you don't fall over your own shortcomings defining the game world etc. is very time consuming. Also, once it's up and running, it's not easy to change something somewhere - you might very easily break the whole system.

In the end, I think that for role playing games (or even just hack-and-slashes) to have decent AI that does not feel like a bunch of schedules requires the game to be two games: Take some sort of dungeon keeper, stronghold or 'anno 1400 (or what it's name was)' game into the background and let it play for itself, thus giving every actor a purpose and a set of needs and actions to be performed. Then put your role-playing game on top of that and let the player run around in this 'world', perhaps even offering him an actually active purpose in the background system: You could have the dungeon keeper master try to hire the hero to let him take care of the other heroes or something like that.

But of course, this also shows that that would basically require two games to be developed at once, which also explains why no one does something like that.

 

[John Yossarian]

In fairness adding that kind of content dramatically increases both content development costs and the cost of testing.

Wasn't this one of your main points for giving RAI a chance though? That using it would make giving NPCs more realistic behavior easier than through scripting? If, with RAI, adding a "he robbed/tried-to-rob you" state to NPCs so they treat you accordingly increases dev. costs so dramatically that it can't be put in the game, how can you say it will eventually be easier and less time consuming than scripting? What is so hard about it that it can't be done (rather than not wanting to be done for cost reasons) today but could be done in the future?

Edit:fixed quotes

 

[Section8]

The NPC steals a loaf of bread, the surrounding AI realise and triggers a message reacting negatively. However, they don't take any lethal action.

Ah, my bad. In any case, it's poor design - if the player picks up an object that has an owner and isn't in stealth mode (regardless of whether they're witnessed), the guards automagically know and punishment is handed down. If an NPC steals in plain sight, there's a "Stop that!" line voiced.

If you're going to be working in and around thse sort of dynamics, it's foolish not to be laying down simple, immutable "laws" of interaction and letting the complexities emerge from them. Regardless of the computational power at your disposal, at either end of that, you have human designers and human end users who need to understand the "laymans terms" of the high level design.

Also, this example is very much indicative of the "for flavour purposes only" design ethic of RAI as a whole. I mean, the whole scenario exists just so the player can witness NPCs reacting to another NPC stealing things. What's the fucking point?

I mean, even if you have a simple mechanic like - NPC steals, guards confront and imprison them for x hours, then you have something that has a dynamic purpose. Suddenly, you introduce potential complexity - let's say the player needs to speak to an imprisoned thief to get some info - then suddenly whatever dynamics you have in place to get into a guarded prison become an extra layer of complexity to an otherwise simple "meet and greet" quest.

Of course, it can potential spiral out of control if say, the imprisoned thief's "income" no longer satisfies the dynamic needs of his family's food intake and so forth, but if you're including that sort of dynamic, then you'd hope it's being leveraged for gameplay purposes and not just there because simulating hunger makes NPC "more realistic".

Then later the AI is in random comment mode and starts up some gossip with the same character. It comes across as very disjointed, doesn't it? Because they are using AI to simulate human interaction, which is problematic because human interaction has a vast array of subtleties. In real life the dude might have been shoved, or spat on, or told to leave the bar or just glared at. The AI doesn't have that and so comes as horribly disjointed. This is why I laughed about the uncanny valley comment. Oblivion isn't even close to uncanny valley.

Surely one of the great strengths of a medieval fantasy world is that you can do away with much "real world" subtlety. Let a witness shout "Guards!" and summon them forth. Regardless of the crime, let the guards beat the accused into submission.

If you want to simulate conversation, then find a way to mask or embrace simplicity (a la Sims speak or Gothic's very generic dialogue). If your conversational model is Greet, Comment, Farewell and a response to each, then question exactly why you're trying to "simulate" social interaction.

In fairness adding that kind of content dramatically increases both content development costs and the cost of testing. And because it won't even come close to human intellect you will still get funny shit happening. The problem is it takes incredible complexity to simulate human social systems and we don't have that. You aren't going to be able to program an AI that a human observer won't notice is fake pretty quickly.

Oh definitely. That's why I'm not really an advocate of "realism" when it's so far out of reach. What developers need to be aiming for is "effective given their purpose." And for Oblivion, it seems clear to me that there's no clear purpose to much of its design, NPC->NPC social interaction especially. Is it there as a hint book? An ego inflator? To make the world seem "more alive"? All of the above?

BTW, I remember in Gothic 3 NPCs reacted pretty pretty harshly to theft too. I generally got killed trying. Didn't seem a particularly superior solution.

Are you sure they didn't just knock you out and steal half your money? It's still an admittedly harsh punishment, but it's just how the Gothic world works.

Oh, and the repetitive dialogue? If i count how many times I heard "It's the same old story" in Gothic 3 it easily equals the Mudcrab lines in Oblivion. It was just less intrusive seeming because it is less memorable a line. Which is probably what Beth should have done but they were AIMING at more realistic human interactions and conversations.

Yeah, that's the problem though - it's a huge step between Gothic/Sims style NPC gossip and something the presents an acceptable standard of "realistic" human interaction. To try and put it in perspective - "back in the day" flat-shaded, untextured 3D graphics weren't really an issue, because the tech wasn't capable of more than that. However, if you were to render incidental cookie cut NPCs in Oblivion as flat-shaded textureless polygons against a multitextured shader 3.0 environment, it's going to be nigh impossible to explain the inconsistency as "artistic license" or "budgetary constraints".

Same goes for a game where you have such remarkably simplistic NPC->NPC conversations against a backdrop of Patrick Stewart and Sean Bean banging on at length about all sorts of shit. It's hard for the player to accept "couldn't have done better" as an excuse.

Also, they shoot themselves in the foot by trying to make every conversation relevant to the player, making it seem even more contrived.

They failed, but like I said, I think they are a company that at least tries, even if the goal eludes their grasp.

My biggest problem with them is two-fold. They don't fucking learn, and nobody gives them reason to. It's embarassing to see the same schoolboy error popping up game after game, and endlessly frustrating that nobody with a sizable readership calls them on their failures.

Well, maybe that's not entirely fair. They do seem to recognise failure of design, but sadly supplant them with a different failure, or something that only "succeeds" from a very subjective standpoint. Oh, and occasionally they grossly overcompensate.

 

[Naked Ninja]

Wasn't this one of your main points for giving RAI a chance though? That using it would make giving NPCs more realistic behavior easier than through scripting? If, with RAI, adding a "he robbed/tried-to-rob you" state to NPCs so they treat you accordingly increases dev. costs so dramatically that it can't be put in the game, how can you say it will eventually be easier and less time consuming than scripting? What is so hard about it that it can't be done (rather than not wanting to be done for cost reasons) today but could be done in the future?

It doesn't decrease the cost of scripting the action. You have to understand, AI uses scripts and animation to perform it's actions. Creating a "spit on NPC" script/animation for a purely scripted game or one more AI driven costs the same to develop. What doesn't cost the same is how the agent makes it's decisions. Think of it like a dialogue tree in an RPG versus sims-speak. In the RPG the designer has to trace every path in the tree and make sure it is inherantly consistent. In sims speak you build up some behaviors and let the AI determine how to use them, and the other AI decide how to respond. The burden of work is lessened for the designer. Spontaneous conversations and consequences arise from the more dynamic system. (The trade off of course is that the conversations are FAR simpler than what a human could generate). However, creating a "waving hands excitedly" animation for use during speech costs the same in development time for both the RPG and the sims game.

Hope that cleared my point up.

Ah, my bad. In any case, it's poor design - if the player picks up an object that has an owner and isn't in stealth mode (regardless of whether they're witnessed), the guards automagically know and punishment is handed down. If an NPC steals in plain sight, there's a "Stop that!" line voiced.

That's true, and indeed, bad design. Still, I think it might be because the AI isn't good enough to escape from prison by itself.

I mean, the whole scenario exists just so the player can witness NPCs reacting to another NPC stealing things. What's the fucking point?

That is the point. Must every character have a detailed purpose? When I play PnP there are NPCs who exist to "come on stage, deliver their lines, then dissapear". Adds flavour. I'd LOVE to see every character be fully detailed, but I understand that isn't likely. In the same way that I understand it when you have buildings that are non-enterable backdrops in other RPGs. They are there to flesh out the illusion of a city. I'd prefer otherwise, but I ain't gonna crucify designers if they put them in.

Surely one of the great strengths of a medieval fantasy world is that you can do away with much "real world" subtlety. Let a witness shout "Guards!" and summon them forth. Regardless of the crime, let the guards beat the accused into submission.

You could definately do that. And they should have. Bad design on their part, but I don't think it is because the AI system lacks the capability. Just wasn't added to it.

And for Oblivion, it seems clear to me that there's no clear purpose to much of its design, NPC->NPC social interaction especially. Is it there as a hint book? An ego inflator? To make the world seem "more alive"? All of the above?

I remember with Morrowind one of the main criticisms leveled against bethesda was that the NPCs were lifeless quest dispensing posts. I think they took that to heart. Like I said, I think they genuinely do try. People also compained about the long "paragraphs of text with html-like topic hotlinks" and viola, next ES game has dramatically simpler dialogue and a topic system with a small amount of traditional branching conversation.

Are you sure they didn't just knock you out and steal half your money? It's still an admittedly harsh punishment, but it's just how the Gothic world works.

They might have. But the point was more that there wasn't any real shades of gray to their system either.

Yeah, that's the problem though - it's a huge step between Gothic/Sims style NPC gossip and something the presents an acceptable standard of "realistic" human interaction. To try and put it in perspective - "back in the day" flat-shaded, untextured 3D graphics weren't really an issue, because the tech wasn't capable of more than that. However, if you were to render incidental cookie cut NPCs in Oblivion as flat-shaded textureless polygons against a multitextured shader 3.0 environment, it's going to be nigh impossible to explain the inconsistency as "artistic license" or "budgetary constraints".

Same goes for a game where you have such remarkably simplistic NPC->NPC conversations against a backdrop of Patrick Stewart and Sean Bean banging on at length about all sorts of shit. It's hard for the player to accept "couldn't have done better" as an excuse.

Mmm, I dunno. Like I said, when I play PnP there are NPCs with deep complicated roles and those who exist for 1 or 2 lines. Same in movies. I don't see it as a problem, so long as the simple NPC is done with skill, in terms of the dialogue. The problem I see is that in Oblivion it wasn't, whereas in Gothic it was.

You could still have :

Greeting -> Comment -> Response -> Bye

so long as it isn't jarring. Which brings me to this point :

Also, they shoot themselves in the foot by trying to make every conversation relevant to the player, making it seem even more contrived.

Yes, it comes across as horribly contrived. Also, the way accents change mid conversation. Make the actual conversations a bit more "backgroundey" and it would work better. In fact Oblivion does have those, some of the simple character greetings work fine.

My biggest problem with them is two-fold. They don't fucking learn, and nobody gives them reason to. It's embarassing to see the same schoolboy error popping up game after game, and endlessly frustrating that nobody with a sizable readership calls them on their failures.

Well, maybe that's not entirely fair. They do seem to recognise failure of design, but sadly supplant them with a different failure, or something that only "succeeds" from a very subjective standpoint. Oh, and occasionally they grossly overcompensate.

Mmm, I don't actually think it is fair, honestly. I think they definitely listened to criticism about their lifeless NPCs and tried to fix it. Their effort wasn't a remarkable success but they probably noticed the criticisms and will try to fix them in the next game. They might introduce new problems, sure, but I don't think it is fair to say they ignore the criticism.

 

[Section8]

That is the point. Must every character have a detailed purpose? When I play PnP there are NPCs who exist to "come on stage, deliver their lines, then dissapear". Adds flavour.

But the near transparency of a CRPG makes this very difficult. Baldur's Gate literally did the "on stage, deliver lines, exit stage" bit, with NPCs hiding behind trees and disappearing, and the cumulative effect was to make the world seem less real. As soon as you lose the luxury of putting things off camera/stage or simply omitting them from a narrative prose, you face a lot of difficulty in putting background characters back into the background.

I'd LOVE to see every character be fully detailed, but I understand that isn't likely. In the same way that I understand it when you have buildings that are non-enterable backdrops in other RPGs. They are there to flesh out the illusion of a city. I'd prefer otherwise, but I ain't gonna crucify designers if they put them in.

The way I see it, the minute you give an actor a speaking role, they're no longer a background extra. Populating a city with NPCs is equivalent to non-enterable buildings. Having NPCs that have three line conversations is like throwing a mostly vacant single room behind every "backdrop" door. Actually, it's even worse, because the player has to actively look behind that door - they're "forced" to overhear Oblivion's wacky NPC gossip.

You could definately do that. And they should have. Bad design on their part, but I don't think it is because the AI system lacks the capability. Just wasn't added to it.

Yeah, and that's what I'm getting at here. The fundamentals of RAI are pretty sound, they're just misused and completely imbalanced.

I remember with Morrowind one of the main criticisms leveled against bethesda was that the NPCs were lifeless quest dispensing posts. I think they took that to heart. Like I said, I think they genuinely do try. People also compained about the long "paragraphs of text with html-like topic hotlinks" and viola, next ES game has dramatically simpler dialogue and a topic system with a small amount of traditional branching conversation.

The thing is, it wasn't much of an improvement. Making those paragraphs shorter didn't make the topic-based dialogue any more like actual conversation, it just made the lengthy monologue delivery more tolerable. It's like the classic "real-time is better than turn-based because you don't have to wait for twenty rats to take their turns." The only "advantage" is that a fundamentally boring design is over quicker.

It's a classic kneejerk response to "public outcry". Like the time our local council worked out there were an average of two-dozen suicides a year where people jumped off the same really fucking high road bridge. The solution? Spend a million dollars putting higher fences along each side so people have to look elsewhere for suicide solutions, or at least take the effort to climb the fences. Now, am I the only thinking that money would be better spent on a horribly neglected mental health system?

Yes, it comes across as horribly contrived. Also, the way accents change mid conversation. Make the actual conversations a bit more "backgroundey" and it would work better. In fact Oblivion does have those, some of the simple character greetings work fine.

I think the Witcher and Assassin's Creed both do a good job of keeping extras in the background, by using crowd VO in areas where crowds are congregated. Though there is a notable fuckup in The Witcher's crowd VO where one of the female actors raises the volume and pitch of her voice (just walk away from it!) above the rest of the crowd, so the loop becomes very obvious.

I guess that's not really an option in Oblivion, where three is literally a crowd, but I think you're right in saying the system itself isn't completely broken, it's just badly scripted. For starters, how often would someone overhear an entire conversation in real life? How often would that conversation involve information relevant to the person overhearing? It's not that far fetched to just feed a stream of fragmented and irrelevant phrases.

Mmm, I don't actually think it is fair, honestly. I think they definitely listened to criticism about their lifeless NPCs and tried to fix it. Their effort wasn't a remarkable success but they probably noticed the criticisms and will try to fix them in the next game. They might introduce new problems, sure, but I don't think it is fair to say they ignore the criticism.

I get the feeling they ignore any complaint without weight of numbers, and that's almost unforgivable. They'll eliminate anything that drew numerous complaints, but the thing is, that's usually the sort of stuff they ought to have picked up on during development or focus group testing. And ultimately, it feels like the changes are being made for reason of popularity, not necssarily to improve the quality of the game.

At the very least, it would probably do them wonders to be even slightly humble in their PR spoogefests. It breeds a complacent culture where sheer confidence blinds you to error. I mean, there's no need to adopt my own brand of hellishly introspective critique and self-doubt, but their end of the spectrum is at least as unhealthy.

 

[galsiah]

That's true, and indeed, bad design. Still, I think it might be because the AI isn't good enough to escape from prison by itself.

That's just rather silly. Clearly the standard way to get out of prison can't be escape. Any coherent system would have escape be a rarity, with serving out a sentence being the common case. In most cases where important NPC X got locked away, the PC would have to break in.
In cases where it made sense for the NPC to break out, they could just break out automatically. There's no need not to include such escapes, simply because they can't be simulated in detail with the player watching. An AI solution is totally unnecessary in this case.

Their effort wasn't a remarkable success but they probably noticed the criticisms and will try to fix them in the next game. They might introduce new problems, sure, but I don't think it is fair to say they ignore the criticism.

Sure, but the frustrating aspect is how they go about "fixing" the problems. All too often, they don't look at root causes, but at symptoms. They then "fix" these symptoms (quite often introducing new problems), yet fail to address the fundamental issues.

For example, in a discussion on TES skill books, a junior Bethesda designer (level designer admittedly), identified this problem:
Problem: players tend to spend quite a bit of time activating every book on a shelf in the hope that he'll find one which grants a skill increase.
Solution idea: highlight the books that give increases so that the player can quickly identify and activate only those.

Directly from symptom to half-baked solution without the question "Is the fundamental mechanic actually a good one?" even getting asked. Personally I think it's an obviously pathetic mechanic, but that's not even the point: questions about the underlying causes need to be asked and addressed. This is one small example, but that type of problem-"solving" is clear throughout TES design.

If you went to a doctor seeping puss, and were given a hair dryer on your first visit, an ice pack on your second, and a portable freezer on your third, would you think: Clearly this is a good doctor - on each occasion he addresses my complaints. The hair dryer to get rid of the puss, the ice pack to sooth the resultant burning, and the portable freezer to address the ice melting.
Somehow I doubt it. Such a doctor would probably be doing his best, but he'd be an incompetent fool nonetheless.


With specific reference to the AI, it's clear that the design is at fault. The attempt at creating a dynamic system is good to see, but the notion that it should be possible to do well without a radically expanded set of NPC-NPC interactions is absurd. Any simulation of a conventional world that allows NPCs to attack and kill each other, but has no system for diplomacy/negotiation, is doomed from the start.

For example:

One character was given a rake and the goal "rake leaves"; another was given a broom and the goal "sweep paths," and this worked smoothly. Then they swapped the items, so that the raker was given a broom and the sweeper was given the rake. In the end, one of them killed the other so he could get the proper item.

What's the problem here? That the wrong guy had the rake? That the system is absurdly brittle? That the violence threshold was too low? That the only solution available to an NPC is ludicrous?
You can tweak/fix your way out of this specific issue, but it's not going to address the root causes. This is entirely predictable without needing to write a line of code - never mind get to testing. If you can't offer an NPC some reasonable means of conflict resolution in 90% of cases, then you can't do dynamic conflict resolution. To think that you can isn't ambitious or laudable - it's stupid.

I also take issue with this:

You're assuming you CAN sit down and map them all. I have programmed AI systems, and my experience is that because of their limitations it becomes a choice of how much independence you want to sacrifice to ensure the system is stable and predictable. Limit it a lot and you ensure it won't freak out, but you limit it's ability to adapt (the scripting end of the system). Give it more free reign and you increase it's capabilities but also increase the chance of it going fruity.

Of course you're right that there'll usually be some tradeoff between freedom and stability - but that's not to say that most situation types can't be anticipated, and designed for. You don't need to map out all the possibilities - you need to anticipate the important classes of possibilities, and adapt to them. So long as you can deduce that e.g. "Where X is an even number, Y is true", you don't need to cover every case of "X is even" individually. With respect to Y, all that matters is that X falls into that huge class.
There's no simple general formula for this kind of qualitative design/balance/anticipation - it takes skill. However, the fact that you can't either do it perfectly or systematically in most interesting cases, doesn't mean that it can't be done at all. You'll always need a lot of testing and tweaking - but by far the most efficient/cheap/versatile place to do that testing is in the initial design (/requirements) phase. Problems anticipated early are hugely simpler and cheaper to fix than those discovered late in development.

Leaving most testing and analysis until late in development is foolish (I'm presuming that they are knowingly not thorough early on - maybe they think they are, but are just crap). A software architect who adopted a "We'll throw some reasonable-sounding ideas together, and test things later..." attitude, would probably get fired. Such an approach to game design is no less sloppy.

Here's hoping they iterate and improve that tech.

Here's hoping they stop with the mindless iteration, and start some rethinking/redesign.