Quote:

Original post by makeshiftwings To simplify things, it's because it's annoying when enemies run away.

I think it depends majorly on the situation of the gameplay. If you have enemies coming at you at a constant rate, you're not going to chase after some lowly AI character. For one thing, a lowly AI character that doesn't stand a chance against you to the point of running for survival shouldn't have any loot that you would be interested in. Killing it should also not be worth much effort in experience.

Another catch for running away would be weapon types. If your character is holding a rifle out in the middle of a barren field, then characters running away are actually more dumb than the ones running to attack. It wouldn't be wise to run unless there's somewhere to retreat safely to.

For this topic, the AI would only be running away if it absolutely knew it didn't stand a chance to survive, no matter what it did. There are a lot of solutions to avoid certain death other than running away. If the tactical situation is suicide, it could change strategies or take cover. If all of it's allies have died violent deaths, it could just hug it's knees and start trembling in the corner, rocking back and forth muttering profanities.

I think believable intelligence in creatures is more important than getting a few experience for stomping on bugs.

well, usually, people running away may drop their weapons, so as to run faster. But they WILL end up, sooner or later, encountering someone they know, and telling them. It would be even better to have them bring in reinforcements... THAt would be Artificial Intelligence...

I believe the way enemies are used by level-designers, so not directly how their AI is written by the programmers, has a huge influence as well. For example, the infinite stream of enemies, often used as a form of pressure to force a player to proceed quickly, makes enemies look cheaper to the player. A few carefully placed, and sometimes scripted, enemies can make a much more intelligent appearance, even though they're not programmed as such.

Yes, enemies getting reinforcements would certainly be a smart move. However, in most more-or-less story-driven, linear games (note that I'm referring to FPS'es mainly), this will make planning the game-flow, the pacing, much more difficult and uncontrollable.

In other words, I certainly believe a bit more intelligent behaviour can be good. But beyond a certain level, the game design itself needs to be changed in order to make it work - in fact, it can become a key part of the gameplay. Which sounds interesting to me, yeah. :)

EDIT: I also think the strive for realistic graphics has something to do with it. The more realistic enemies look, the more realistic behaviour we expect, or even demand, them to have. This may not at all be something that works with the core gameplay. Which makes me believe that some of these games had better not adapted that overly realistic look.

AI in games definately needs to be pushed futher, there is plenty of room for improvement. However, just adding "fear" without a solid plan for enhancing instead of harming gameplay, would be a bad idea. Scared enemies may already know that they are going to die. There are many possilbe responses to fear:

1) Becoming more aggressive: hitting harder, kamikaze techniques, hitting more often
2) Becoming more cautious: dodging, blocking, defending, sneaking, regrouping, staying just out of range, spending more effort seeking positions of advantage (archers flee to towers or fortes for example)
3) Becoming more jumpy: loss of accuracy in attack paired with increase aggression in attacking

It is possible to add fear to the enemies without necessarily making the game easier. Fear done right could even make it harder.

Quote:

Original post by T1Oracle However, just adding "fear" without a solid plan for enhancing instead of harming gameplay, would be a bad idea.

Imagine the player standing on one side of a pit, and enemies running toward him from the other side. These enemies are so dumb, they just fall right into the pit as they run at the player. This is no different than enemies who run directly toward the player with a club as he aims a shotgun at their head. The same amount of suicidal behavior. I don't think doing something to stop this type of behavior is ever a bad idea.

Most designers and programmers probably note that by having that first idiot get his head blasted off, the others are able to get in free hits, and there for the difficulty and challenge is worthy, and the AI is competent. But they're failing to realize that real creatures don't behave as a single unit in this way. Not unless a large population of their kind is under threat, or, for really selfless creatures, when failure instead of victory will result in horrible consequences for many others. Most of the time, this is not the case in video games. Usually, monsters want food, and bad guys are nothing but lackeys being paid by the real bad guys, both of which who would not risk spend individual lives to get an upper hand in combat.

Having the AI operate as a single unit under normal situations, especially human AI, results in them appearing to be very fake entities. A clone army. It is as if the AI itself is using the fact that it is cheap to produce as an advantage in combat.

You should consider the goal of your AI, before implementing fear behaviour simply because that's what happens in reality. After all, games are virtual environments, with rules that do not necessarily have to map to the ones in reality. While the representation often resembles reality, underlying rules don't have to. I think your whole issue is rooted in this 'representation/game rule' inconsistency rather than a gameplay flaw. Go with a cartoony style, with large-eyed, two-teeth-sticking-out monsters and suicidal en-masse attacks suddenly feel a whole lot better-fitting.

I agree that fear behaviour can allow pretty interesting situations, and it's definitely worth exploring. On the other hand, I can see various issues attached to it, as I already stated in a post above. This sort of behaviour has its cost, it's not an instant improvement, not even necessarily an improvement at all. On the other hand, the situation right now is probably too quantity-over-quality oriented, so perhaps it's about finding a balance somewhere rather than totally going to one side of the spectrum.

*shrug*

Players lack this same fear (in-game, of course). It's part of the reason we play games.

Quote:

Original post by CadetUmfer *shrug* Players lack this same fear (in-game, of course). It's part of the reason we play games.

I don't agree at all. We should fear our characters dying. If the game is designed well.

Quote:

Original post by Captain P You should consider the goal of your AI, before implementing fear behaviour simply because that's what happens in reality.

My goal is to have believable I in my AI. It doesn't really matter to me if the characters are ninja turtles or army soldiers. I want to sell the characters to the player as believable living things. That's just the way I am.

Quote:

After all, games are virtual environments, with rules that do not necessarily have to map to the ones in reality. While the representation often resembles reality, underlying rules don't have to.

You're absolutely correct. The whole fear of death issue shouldn't map to games whose AI are not afraid to die, or who don't really consider death 'dying'.

Quote:

I think your whole issue is rooted in this 'representation/game rule' inconsistency rather than a gameplay flaw. Go with a cartoony style, with large-eyed, two-teeth-sticking-out monsters and suicidal en-masse attacks suddenly feel a whole lot better-fitting.

Two-teeth-sticking-out monsters are supposed to be suicidally dumb. Or enraged enough to not be capable of thinking properly. But the representation doesn't matter much to me when I'm playing games. Only to the point where I realize the development process was seriously neglected and that the game is supposed to be simple. In other words, I want believable creature enemies in both Zelda and Half Life. But that's just me.

Quote:

I agree that fear behaviour can allow pretty interesting situations, and it's definitely worth exploring.

Jumping into certain death would not show a lack of fear. It would show how incompetent or fake AI is. Not fake as in unrealistic, but fake as in not making any sense. Even robots and stickmen should be programmed with directives to avoid certain death.

Quote:

On the other hand, I can see various issues attached to it, as I already stated in a post above. This sort of behaviour has its cost, it's not an instant improvement, not even necessarily an improvement at all.

It is indeed an improvement. If 50% of your players get drawn into your game world more because your AI react in such an understandable way, then I say it's worth all of the small costs. If there are big costs, then that's what I would like to hear about. It looked to me like you only had concerns with the reinforcements issue, which there are many ways around.

Quote:

On the other hand, the situation right now is probably too quantity-over-quality oriented, so perhaps it's about finding a balance somewhere rather than totally going to one side of the spectrum.

I would think that having your AI react to a no-win situation is always going to be a great thing. I honestly can't think of many games that I've played where this problem wasn't subconsciously bouncing back and forth in my brain, except when the game itself has limited functionality. I can't count the number of times that I've had the impression of the AI being a single unit. I'm fighting against IT. The game world god. Not them. Not the globins. Not that one guy. IT. There are some games that do very well to prevent this type of feeling, but not the majority. I am very interested in hearing about drawbacks to avoiding this. Hearing about the costs was actually my reason for posting.

People that are trying Dungeon Crawl often are finding themselves moving backward quickly, and frequently check if something don't block theirs retreat path. The problem is when something would appear in theirs path. Then characters would show wonders with attempt of surviving.

Yes Dungeon Crawl monsters have fear, and it could be a big problem to finish some off because they have tough skin. So if they retreat, they would return when they'd regenerate.

One of the costs of intelligent behaviour like calling in reinforcements is the added burden for the level-designer. The games pace becomes less predictable, and therefor harder to direct by the level-designers.
Fearfull behaviour, besides adding programming hours, has a similar, although probably less significant, impact. In the end, it'll change the way the game plays, and can do so to a level that changes the whole games design, it's very nature. That's why I believe it shouldn't be added without much further thought. The effects can make the gameplay different in unforeseen ways.
My remarks are probably more of a caution sign - think about, test it and determine if it fits the games style and gameplay - than a reason not to implement it.

As for simple gameplay, that's not always a case of no I in the AI. It can be overall lazyness on the design part, or inexperiencedness, lack of polish... implementing fearfull behaviour can spice things up, but it can be implemented in a lazy, unbalancing way just as well. I've seen the Half-Life enemies you liked been used in absolutely horrible ways - how many custom maps don't feature grunts around a corner, looking as if they were just waiting for you to show up to get killed? With no cover around to show their more intelligent behaviour?

As for immersiveness, that's a good reason to introduce this aspect. As I said before, there are certainly reasons to go for this. But not every game will benefit from it, it won't fit every gameplay style. Which is why I believe it's not a magical improvement. In certain contexts, yes, in others, no. But for those shooters that constantly up their visuals, it's probably a good thing to have - their visuals have become too inconsistent with their characters behaviour, breaking immersion and reducing the fun. Another solution for them would be to adapt a more abstract or iconic style, such as Team Fortress 2 does. I believe that's a good example of adapting the visuals to the gameplay: a fun, not-too-serious gameplay with an equally not-too-serious look. I'd say, it's a pretty inviting game. :)