Interesting AI for battles

[Hormalakh]

I'm finishing up IWD1 here and one of the interesting bits in that game is the

story of Marketh and his abuse of the drow girl, Ginafae. Or Dirty Llew's misogyny. One of the things that you learn is that they're abusive towards women and this is part of their personality.

 

 

It woiuld be interesting if some of the harder bosses/demi-bosses/special characters had AI's that were completely different than the other monsters and these AIs matched their personalities. So that players who are paying attention to the story and that pick up on these things can use them to their advantage during battles.

 

For example, maybe Marketh/Dirty Llew could target women in the party first

implying his hate of women or whatever

and wouldn't attack men until all the women are killed. Or for example, since elves hate drow and vice versa, the elves/drow would first attack each other in addition to other considerations made during the battle.

 

Maybe Malavon

the drow wizard with the blinded gnomes at Lower Dorn's Deep in IWD1

would attack gnomes in the party until they're all dead and then move on.

 

I think this would definitely make the story more interesting, make paying attention to enemies' weaknesses/personalities more interesting, and make battles more interesting because of the strategy involved.

 

AIs that have prejudices towards mages, women, races, cultures, etc would be very cool. This would make the mechanics match the story, and if you can avoid the loopholes that players will find (make the AI chase one character around only while the others shoot him down) this can be more enjoyable.

 

What does everyone else think?

Edited March 30, 2014 by Hormalakh

[Ffordesoon]

Would be awesome. No idea if we'll see it, seeing as this screams "out of scope" to me, but it would be nice. Certainly, an SCS-style mod that makes all enemies act "narrative-friendly" would be appreciated.

[Hormalakh]

I don't think 5 or 6 lines of extra code for maybe a handful of colorful NPCs is out of scope. I'd say this fits within the core three golas that Obsidian is trying to achieve.

[Elerond]

Behavioral weaknesses in enemy tactics sounds like it could be tactically interesting, especially when you add phobias in the mix, like for example enemy in fire proof/resistant armor that fears fire, that would allow you to block its path or drive it back with spells and items that do fire damage, even though they do little/no harm to it.

 

But I also fear that it would be out of the scope, because even changing attack priorities need much more work than adding 5 or 6 lines of code for behavioral script, because it needs to be tested with multiple party combinations to see that it actually does what it should and that it don't break when there is for example whole party of characters that fulfill new priority scheme, which means that characters should use normal attack priority, etc. things that at end will probably mean that there need to be quite much additional exception handling in script, meaning that you probably need more than 5 or 6 lines to make it so that AI don't break in any situation.

[Jarmo]

Differing AI's for different enemies is something I've advocated before.

Not based on personalities as such, though it could bring additional color to fights.

 

In additon to ones mentioned, there could be multiple possibilities.

 

- Ignores women, because they're of no consequence.

- Hates wizards, attacks them first.

- Respects wizards, attacks them first.

- Tries to take down the biggest fighter first.

 

But the thing I've mentioned before, would be differing quality of AI's based on tactical prowess of the enemy.

 

- Something like zombie or troll would just attack whoever is nearest.

- An automaton might always switch target to whoever did the latest damage.

- A party of dark elf commandos would use the best tactics the AI can offer.

- A party of adventurers might have a couple of clever guys and a barbarian who just rushes in.

- An evil spellcaster with delusions of grandeur might give no heed to whether his fireballs hurt his companions as well as intended targets.

[Hassat Hunter]

- Hates wizards, attacks them first.

- Respects wizards, attacks them first.

So, everyone just attacks Wizards?

[Jarmo]

 

- Hates wizards, attacks them first.

- Respects wizards, attacks them first.

So, everyone just attacks Wizards?

 

 

Trying to illustrate there can be multiple reasons to do that.

There could also be someone with nothing but contempt for all spellcasters, ignoring them and their pink sparkle at his peril.

[anameforobsidian]

There's no reason for that to be listed in describing the AI though, just an attacks <variable> first would provide enough characterization through mechanics.  Although, honestly forcing npcs to attack one person would probably be needlessly limiting AI responsiveness.  What if they just had aggro counters and <variable> (class, gender, ethnicity), would have a static modifier?  That way, even if an NPC doesn't hit women, they might change their mind if a woman is about to kill them.

Edited March 31, 2014 by anameforobsidian

[Messier-31]

- Ignores women, because they're of no consequence.

 

Biggest weakness: 100% feminine party, snicker

[Frenzy-kun]

This time I agree with you. Bosses should be something more than a harder combat. In every game, bosses suppose a change in the standard gameplay to offer new gameplay value.

 

But I would improve the idea a bit, and add new elements that cannot be found in regular gameplay phases. Since this game has combat rules and people usually dislikes to break the rules in order  to give bosses new abilities, the concept of arena could be easily exploited. An arena prepared for the combat. For instance,  some archers over an unreachable places covering the main enemy. Magical barriers that requires of strategic movements in order to deactivate. Some areas in the floor where anyone over it gets healed, having a character to make a ritual and preventing enemies to step in the magic circle. Well endless situations.

 

I think personally that they overdid with sendai in TOB, but that could be also an example of what could be a nice boss. It could be improved by letting the player to destroy the statues.

[Pipyui]

As Ffordesoon said, this business feels too out of scope to me; but besides that, I find it hard to imagine that any enemy would allow their biases to so affect combat tactics unless they were particularly confident in their own capability, or otherwise really, really hung up on something (like super-crazy-misogyny).

 

I'd prefer to see more generalized combat AI archetypes or personalities such as Jarmo addressed:

 

 

- Something like zombie or troll would just attack whoever is nearest.

- An automaton might always switch target to whoever did the latest damage.

- A party of dark elf commandos would use the best tactics the AI can offer.

- A party of adventurers might have a couple of clever guys and a barbarian who just rushes in.

- An evil spellcaster with delusions of grandeur might give no heed to whether his fireballs hurt his companions as well as intended targets.

 

Monsters can and probably should be fairly tactics-locked in this context, but human enemies could each be granted one of several AI script sets so that you might see aggressive, elusive, runs_when_alone, opportunisitc type enemies.

 

And of course bosses would display unique tactical behaviors, but I don't feel that these need tie to the more particular facets of their personalities, just the overtones.

[Jarrakul]

I think this whole idea is really, really cool, although I too wonder about scope. That said, the devs know about scope issues better than we do, so I'll just say I love this idea.

 

One thing I'd really like to see is something where, if enemy A takes damage, the attacker generates aggro from enemy B. In other words, one enemy could be protective of another enemy. A pair of lovers, for example, might each get really angry when you hurt the other one. Or a commander who cares for his men might do his best to protect them, while a necromancer's minions might do everything in their power to keep their master safe.

[Hormalakh]

Very cool ideas Jarrakul! Those would be very interesting fights.

[JFSOCC]

Good AI is fairly predictable. You know that when you find a certain type of enemy, he is going to play to a certain strategy. this allows you to learn from and prepare for the next time you meet a similar enemy.

 

Having Boss AI's be different makes the mix all the more interesting, especially since Bosses are one offs anyway.

 

I don't want to make assumptions of scope (though yes, sometimes I do) as I simply do not know enough about the subject to know how much effort is required to put certain features in the game. I think we would all be wiser if we didn't make these assumptions. Especially because it makes it sound like you don't want to defend your good ideas. "I know this is probably too much, but" isn't exactly inspiring anyone. Besides, a good idea is a good idea, and whether it fits the scope of the game is a decision you don't have to make.