Artificial Intelligence and Fake Difficulty

[soulmata]

Short preface: I've long been a die-hard fan of many BioWare, Black Isle, Interplay and Obsidian endeavors. Thousands of hours wittled away while I stomped through Durlag's Tower, assassinated Yxunomei and developed a crush on Stale Mary. I've even got a tattoo of the rune of torment. True story. So, in my wildest fantasy, Project Eternity produces a piece of work with epic gameplay from start to finish, telling me a story so complex I need Ritalin just to sit down and comprehend it, offers me a combat experience so intense I swear I feel my characters bleeding out and gives me aesthetic engagement so rich I get vertigo just looking at the splendor before me.

 

I can already tell PE's going to have an amazing story, an engagine engine and probably a beautiful soundtrack. Out of that, there are two components I'm interested in most that I haven't heard much about - artificial intelligence and 'fake difficulty'. As PE is at this point mostly player-backed, I have a lot of faith, but also hope I can get at least a small word in the developers ears. I'd like to hear their plans for PE when it comes to these two facets. Now the details:

 

-Artifical Intelligence-

AI is the crux of most games, because of the compounding effect it has on replayability. Many advertising campaigns go out of their way to tout how deep and sophiscated the AI is (use F.E.A.R as an example), but players are often left disappointed. Sometimes that's our own fault - if the AI can be abused, it will be abused, and as a software engineer myself, I know there's only so far you can think ahead. Details so far on what sort of attention the AI's going to get in PE have been sparse, so I'd like to hear from the PE team on what some of the plans are, if any at this point.

 

Some questions I think about (I don't expect answers, they are just things I like to think about):

Will the AI receive at least as much attention as, say, music and art components? This is in terms of raw hours of development.

Will it be designed with varying degrees of difficulty in mind? Adapting to the player's skill is an example.

Will there be an entropy system in place designed to ensure (or at least attempt to ensure) a player never finds the "golden rule" of Player vs. The World that always favors the player? Examples are an AI'S chance to do something completely stupid and/or unexpected, just to catch the player off-guard.

Is it being modeled after older Infinity engine games, or is an entirely new approach being taken?

 

I'm very interested in hearing what the developers want to do with AI.

 

-Fake Difficulty-

I really do like the difficulty-enhancing options mentioned in previous updates, such as making creatures do more damage and giving the player only a single savegame (I actually maintain a site dedicated to playing games on a single run, and Baldur's Gate was the first I catered to). Some of the options fall under the scope of fake difficulty, such as increasing/decreasing the amount of damage a player receives/gives. Those aren't neccessarily bad - they can add an element of danger where it might have been a cakewalk before - but I'm wondering if other aspects of difficulty enhancement are being considered with them. Examples:

 

Rather than just making a particular spawn point have every possible low and high level enemy spawn, how about also making those same enemies be much, much more aggressive? They'll focus on killing as fast as possible, rather than being elegant.

Rather than spawning and sitting, waiting like cattle for the player to come and slaughter them, how about making them actively hunt the player down once the player is in "their dungeon" or in "their territory"?

Rather than having quests and other objectives which are aimed solely at the player, how about competing adventurer groups which might come claim a reward, honorably or dishonorably, or perhaps come after you in revenge for stealing their bounty?

 

I don't expect answers on many, or even any, of these, but it is something I hope others on this forum are interested in discussing. What are your opinions on engineering a solid AI? What are you expectactions from an epic game? How heavily does AI factor into replayability for you?

[Radres]

In BG2 there wasn't much of an AI, i think for most of the RPG's it is an aftertought. And as a consequence the tough combats in BG2 had cheap tricks like the instant buffs some wizards and monsters received. I hope this will be different but i'm not betting on it.

[soulmata]

And it's concerns like that we should voice now. Sometimes the AI has to rely on trickery - there are aspects of the human experience that it just can't emulate - but there are other cases where it's obvious the developers simply didn't have the time or inclination to make the AI strong. BG is a great example. The vanilla AI in BG was pretty weak and quite easily abused. However, after-the-fact improvements, such as BG1TuTu and Sword Coast Stratagems, demonstrated that the AI was not nearly as constrained by the engine as some believed. SCS made the AI much more challenging, and did such largely within the confines of the game - by making casters focus on disabling your party more rapidly and making the AI much more aggressive about using potions, et cetera.

 

Any time the AI can play by the same rules as the player and still win occasionally is a great thing. Any time you are playing and actually believe, if only for a minute, that you're competing against an intelligent agent working in direct opposition to you, the developers deserve a round of applause, because that's incredibly hard to pull off. And it's why I'm really, really hoping PE takes the time to make it happen.

Edited October 3, 2012 by soulmata

[Osvir]

I think my own personal dream AI is somewhere along your suggestions/ideas soulmata.

 

Where you enter a bar, and depending on your alignment/reputation, the thugs of the place might even come up to you and scuff you around "Ya think ya are something huh?" kind of thing, or where they might suddenly attack you just because they are drunk and violent *shrug* a living AI

[soulmata]

A convincing and competent AI is a challenge. However, too many games today cater to the lowest common denominator - if the game is pretty, relatively bug-free and has good music, AI can be shuffled into the trash and all opposition can resort to a suicide march against the player whenever spotted.

 

I want PE to go beyond that - to consider that for a story to be truly compelling, there have to be elements of danger and surprise. There has to be incentive to plan your future battles, and there needs to be a reason to think on your feet, rapidly. And those pressures shouldn't just exist solely in the temporary space of an individual encounter, but should spread throughout the game. Take too long trying to find a cure for a disease plaguing a town? Well, most of that town's population is now dead - or perhaps someone else found the cure, and the reward goes to them.

[Parallax]

Soulmata, are you sure you're not telepathic? Because I've been thinking about pretty much the same things, even right down to the comparison with SCS and feeling like you're actually playing against another person, and not just a bunch of poorly scripted punchbags. Enemies working in a team, being aggressive and pressing the advantage, using spell combinations as a player would...those kind of things made for the most intense and memorable clashes in the games. Having difficulty relate to how cohesive, smart and aggressive the enemy are, as opposed to how hard they hit, how many spawn or how much weaker you are would be ideal.

[soulmata]

If I look over the history of my own gaming experiences and try to isolate those moments which truly made a game memorable, I find most of them come down to a small handful of categories:

 

1) Some plot hook or compelling story element that made me want to keep playing just to see how it ends. (P:T comes to mind)

2) A combat system that made me strive to improve my tactics, rather than just iterate through every combination of tactics to see which broke the AI the worst. (X-Com comes to mind)

3) A sufficiently dynamic world that made it feel alive, but not random and repetitive. (Heavily-modded Skyrim comes to mind).

 

Aesthetics never seem to be something that makes me want to play a game again, nor music or even character customization. It ends up being how much the game makes me think, and how much it forces me to react, rather than me fully dictacting the course and pace of the game, and expecting it to react to me. That's why I want, very badly, to see AI get a decent focus in PE. Because that's where the hook will be - that's what will keep you coming back, that's what will convince you to play again and again, to buy DLC or expansions, to spread the word to your friends, to stay up for 36 hours straight because you can't put the game down.

[metiman]

Maybe they should hire DavidW to test their AI scripting for them.

[soulmata]

The thing is, a lot of good AI can be developed through mimicking. If the player performs a behavior over and over, it's likely a successful behavior - one the AI should try to emulate. It can sometimes break suspension of disbelief, but other times can be incredible. Look at Unreal Tournament, and the ability to tell AI bots to bunny hop. Completely ridiculous for an AI to do, yet it was much more effective at making it feel like the bot was an actual player - because it was doing something absurd and unpredictable.

[Jarmo]

I'd like them to make the best AI they can to make the opponents act believably, and use that on all difficulty levels.

 

Wouldn't like passive enemies at lower difficulty, nor would I like to see mindless zombies coordinate with military precision on higher ones.

I'd like goblins or orcs rush in mad barbarian rage, dark elves trying to trick or ambush and adventure parties fight like they know what they're doing.

 

If there are different AI styles, use them for different opponents, not for different difficulty levels.

[ledroc]

I LOVE this post and have been thinking the same things for years. Most recent games are focusing on the wrong things in my opinion. Sure the flash gets the games off the shelves aka Skyrim but the AI stops people from playing them, aka Dragons in Skyrim

 

I truly hope this game gives us a legitimate challenge bordering on frustrating but makes it fun at the same time.

 

Best table top D&D session I ever played pushed the party to the limit but still let them win. Face it, none of us want save game spam fest, but we also dont want to roll over everything.

One of the big pieces of this is caster viability. How many spells did you use in the BG or ID games? Very few in my experience. I loaded my party up with bows and went out sniping.

This was fun at first but got a bit boring later on. I kept seeing all these fun spells in my spellbook that I had little to no use for.

 

This was due to lack luster AI and relative monster strength.

Kiting is a good example of exploiting the AI. There are several encounters in the above mentioned games where you could stand behind a tree and the AI couldn't path to you so the monster would just go left, then you move, monster goes right, you move. All the while your party is peppering the poor guy with arrows.

 

I am really also hoping for an "aggro" based combat model, so much so I think I will put up a poll about it.

[metiman]

Well at least admit that an aggro based combat model is an MMO combat mechanic if you do put up a poll about it.

[soulmata]

I'd like them to make the best AI they can to make the opponents act believably, and use that on all difficulty levels.

 

Wouldn't like passive enemies at lower difficulty, nor would I like to see mindless zombies coordinate with military precision on higher ones.

I'd like goblins or orcs rush in mad barbarian rage, dark elves trying to trick or ambush and adventure parties fight like they know what they're doing.

 

If there are different AI styles, use them for different opponents, not for different difficulty levels.

 

Something as simple as this actually goes a long way toward keeping suspension of disbelief alive and adding replayability to games. When animals behave like animals - skittish and flighty. When humanoids behave like humanoids - territorial but not suicidal. When a group of enemies heavy in ranged gear but weak in close combat take advantage of this and try to kite the player.