Lo-Fi Games

The Next Level

Now hes even more human. He knows he will die if he tries to save you, knows that every line of reasoning tells him to run away, and that he cannot help you. But he fights anyway, to the death. Out of loyalty, out of love.

This has probably already been thought of but here are my two cents about a higher level of AI psychology / sociology.

It would be interesting if your AI companions would be more or less likely to stick around in a battle depending on their loyalty, which in turn depended on how well you treated them (bandaging them, taking them to the doctor, protecting them, pulling them out of battle when they collapse.) This would in turn mean that fresh recruits would be unlikely to stick around and die for you, while your seasoned veterans, that you have worked hard to train and keep alive, would be more than willing to take on 1:5 odds together with you.

The Next Level

In addition to this you could also consider more individual differences between AI characters. Put the level of individual intelligence, bravery, loyalty (to money or to a person), and other characteristics (that might vary between characters) into the equation.

A dynamic perspective?

And of course to make the game dynamic and less revolving around yourself, your AI companions might have relations to one another. Lets say one member of your elite crew has been protecting and medically tending to a fairly new recruit. The recruit might want to stick around in an especially deadly fight to keep his friend alive, if the elite member in turn collapsed, this recruit would be more than willing to pull your elite member out of the battle and desert you.

Another layer to the dynamic perspective

Lets say the recruit and the elite both make it out alive, the recruit tends to the elites immediate wounds before taking him to a city, spending his hard-earned pay on a doctor to mend the elite further. When the elite has recovered, he then proceeds to convince the recruit that the player is in fact an honorable leader, and they both end up back at your doorstep or in your camp eventually.

In conclusion
A lot of make-believe and wishful thinking in here, but hopefully (sometime much further into development) at least some sort of dynamic perspective could be introduced.

P.S Dunno if this goes in gameplay or general, move as needed.

What you are suggesting is good, but quite complicated and very hard to implement.
Personal stats like bravity and others are easy to track and may be nice addition.
But story you wrote here will require hundreds of hours on script writing. Maybe it's better to simplify it somehow? Like relationships between fighters of one squad are slow getting better with time or drop if someone of them runs like a coward. And good/bad reationships in squad may give bonuses/penalties to it.

I realize that most of it would be difficult to implement, but the idea was that it should be implemented in a dynamic fashion. You could make a relation value between AI in the same squad (If you're feeling ambitious, also between AI in different squads), that changes depending on actions such as medical attention, carrying incapacitated person out of combat etc. Depending on the size of this value you could have different sets of AI behaviors that would kick in.

I realize writing AI is a lot of work, and I wouldn't mind seeing something a lot simpler introduced into the game within the next few months, most of these suggestions were just things I'd like to see in a finished game someday.

characters will definately have personal relation factors that affect their descisions. So given a choice between saving two squadmates, the character will be more likely to choose the one he likes the most. The relation will be increased by time spent together, based on jobs (eg bodyguarding, teaching) and past actions (eg rescuing)

That sounds great Chris! I look forward to watching this game develop.

Chris_The_Great wrote:characters will definately have personal relation factors that affect their descisions. So given a choice between saving two squadmates, the character will be more likely to choose the one he likes the most. The relation will be increased by time spent together, based on jobs (eg bodyguarding, teaching) and past actions (eg rescuing)

Hi Chris_The_Great !

Will we as the consumers have full access to the scripts ?

Is there going to be any content locked from us at all ?

I know some developers don't like people messing with particular parts of their work,which is understandable.

Temper

hmm... so under the time together concept would the units assigned to guarding your base be less loyal than your traveling companions? Basically, is the "time together" the amount of time the unit was in your employ or the amount of time the unit was physically with your character? If it is the latter, would the guards at your base be hesitant to protect you or have a higher chance of feeling during an assault on your base with you there (just visiting)?

I think it would be both.
You would gain positive relation faster if you were with them physically rather than just having them guard somewhere. It would work the same way for losing positive relation as well.

With that being said, if you lose a battle and it hurts your relations with someone, it would have less of an effect on those you hired, as they aren't there personally. And only heard about it as opposed to having lived it.

This could go the same for heroic/evil deeds.


There definitely need to be 2 separate stats for this. Friendliness/Respect and Trust/Belief.

For a rough Idea, Friendliness/Respect would be how much they like you.
Whereas Trust/Belief would be how much they agree with you.

I can more in-depth but I'm tired and need to sleep, it's 2 A.M for me. I'll elaborate when I wake up.

Interesting ideas Scab,

I specifically like the idea of your troops having individual relationships, though the prospect of AI couples counseling or dispute mediation, gives me pause.

That said if your troops have individual relationships, how would the men with you, in the thick of it, if you will, regard the ones say patrolling a base?

Would the elites you've had since the beginning be regarded with awe by the possibly newer guards?
Do the older troops with you look at the ones in the base as unproven or even young/green?
Would a veteran be troubled by having to rely on an untested recruit to watch his back?
Does the recruit feel compelled to prove himself among such an august group of veterans?

I know such ideas are potentially too abstract for the purposes of a game, but they're still interesting ideas.

The questions you are raising are mainly sociological Varg, and they are quite interesting to ponder upon indeed. I believe that maybe in 20 years when we have processors that are 10-100 times as good as they are today, implementing psychological and sociological depth into AI will be more than viable, as long as we have programmers who are still interested in breaking the boundaries of game development.

And about posted guards at your base, I'd say if you wanted loyal and strong troops you'd train recruits to elites and then post them at your base. Seems logical to me that a fresh recruit, who is only guarding your base because you're paying them too, who has never even seen your face, wouldn't even think twice before running when a large squadron of enemies appear.

Scab wrote:I believe that maybe in 20 years when we have processors that are 10-100 times as good as they are today, implementing psychological and sociological depth into AI will be more than viable

Uhm....We can do that already.....It's just a simple matter of multiple relationship values/simple if-then statements/triggers that your character does.

The reason it's not been done yet.....No one cares to/it doesn't fit into most games.

How many epic squad based games are there?


If you want psychological and socialogical depth like that, look at dwarf fortress, you'll find it.

Synyster, Dwarf Fortress does have INCREDIBLE individual depth for the dwarves, but I would say that the sociological department is a bit lacking. I mean people get pissed when their friends die but that's really the only phenomena I can think of.

But Dwarf Fortress pretty much proves that the processing capabilities do not exist today (unless somebody comes up with really REALLY good code), the more cluttered and the bigger your fortress gets the slower time progresses, in a fortress with more than 100 dwarves you're usually gonna be down to probably half the original non-laggy speed.
On top of this the game also has quite a few bugs in both the military and medical system, not to mention an enormous amount of crashing in most games that go past 5-10 years.

You might be right that it could be possible to add an incredibly simplified sociological part to Kenshi, but even if you boil it down to just a few values we're still talking about a LOT of AI relationships that need to be handled. In the end the only viable way to give AI depth today and still keep performance high is to come up with great and effective ways to handle the processing.

My original point was, that if you had for example a processing capability 20 times of todays processing power, making a dynamic, living world filled to the brim with thousands of AI, that all have their own neat psychological and sociological calculations, would be much more viable than it is today.