TIGSource Forums

Hello TIG,

Quite a long time ago I started a topic about my plans to make machine learning and character development software! Well, they're both very nearly complete!

Link to that topic here: http://forums.tigsource.com/index.php?topic=29465.0

To demonstrate some of the possibilities for my software I've made a video on modelling moral decisions in Unity3d using my software coAdjoint Orbit. By using proper artificial intelligence I (hopefully) demonstrate that it's possible to make much more interesting moral decision systems than we see in current games.

http://www.youtube.com/watch?v=GYPgMeVy2lI (http://www.youtube.com/watch?v=GYPgMeVy2lI)

If you enjoy the video or want to discuss my software I'm very interested to hear!

Cheers,

Tom

Very interesting videos and the system itself. As for me, I'd feel more comfortable with bars with different width rather than colored circles (let's just ignore the fact, it's for Unity and I'm not a Unity man at all :)), but that's just me. Looks promising.

Machine learning ftw. :giggle: :handthumbsupR:

Thanks for the interest!

Just as a side-note, although you have to train Orbit in Unity, your training and network are compiled to a C# .dll. So if you can interop with C# then you can use it in your games as well.

I tried to do something like this before, in a different way.  You can see the results of my experiments here: http://spaghettilogic.org/friendship_game

I eventually stopped working on it because it makes playing the game a lot more complicated.  Part of the reason games are fun is because they're not morally ambiguous.  You don't have to feel bad about blasting zombies & trolls. 

That being said, I think there's a really interesting game to be made if someone can come up with a less overwrought system than I did.  William Faulkner said that what makes a story interesting is "the human heart in conflict with itself".  Like in Serpico and Macbeth.

Also, when it comes to interactive storylines, you might want to take some advice from an expert: http://jeff-vogel.blogspot.de/2012/10/what-makes-choice-interesting.html

I think this kind of systems is not for "Kill 5 trolls and gain +5 to aggressiveness for each of them", it's more about "Kill a troll and save the village / Do not kill a troll and let him destroy the village / Join a troll, destroy the village, eat villagers' hearts, kill a troll, make an amulet from his leg", it's about decisions you make through the game. Just like Mass Effect, but more complex and with all shades of good and bad, not only Paragon/Neutral/Renegade.

Those who play ME will agree that it was always fun to use Renegade interrupts in ME2/ME3, but why do I become a worse person after I've killed a bandit or corrupt something to make the upcoming battle much easier? Okay, give me points to "aggressive" and "selfishness" and let me go, I am a Hero of the Galaxy, remember? :giggle:

But in the end, it all depends on the game designer and how he integrates this system into the game, of course. And if the game's core idea is "kill everyone and don't think about anything", then, I think, it'll be strange to use morality system in it. :)

I use Java for now, but anyway, it's nice to hear that. :gentleman:

0x0961h, you got everything spot on  ;) That's the point of this demo anyway.

Just as another note, when I was making the software I didn't even have morality systems in mind. I was thinking more about adaptive A.I. - you design a network of skills that progress through some character development system in game, you train Orbit to understand what levels tell you about the playstyle of the user and then use those classifications to change the behaviour of the A.I. This demo is to show the versatility of the software.