Related to a previous post ( What leads to better User Experience? ), regrading what would lead to a better user experience, I’ve decided to follow up on that information by focusing around what experts think would help create a more immersive game or user experience when it comes to creating an AI.
Kimberly Voll – Game Designer, software engineer and cognitive scientist
Kimberly Voll during a conference for GDC, discusses how a simple AI is better than a complex AI. When designing your AI, to start stupid, making it follow the player as the first step for example. Then to polish from there, adding new features as you go, and to resist complicating your AI. The AI you are creating should provide a challenge or help the immersion of the game, think simple, work with your brain.
She also talks about the behaviors of the AI, choosing specifically a behavior to work around, is better than having a complex model. I think I can see where she is coming from, having a simple AI that only has a select few amount of jobs is better than having it try and do everything within your game. Bottom line, having a simple AI can work better than a complex one.
My supervisor Craig also said “Instead of having one AI trying to do everything, to split them between multiple AI’s”, this is a good example of taking simple into consideration.
James Portnow ( Introduction to James Portnow ), a professional in the game industry does a short 8min video talking about AI in video games ( Game AI – Funtelligence – Extra Credits ). He also dispels the misconception about complex AI not being good AI, “Game AI’s are not going to do their job better just because they are more complex”, he also talks about the purpose of a Game AI and how a learnable AI works and intuitable AI.
But he also makes a lot of great points when it comes to the complexity of a Game AI, well basically just putting down the idea of complex AI’s.
- Don’t use enormous behavior trees when a basic script could get the job done just fine
- Don’t try to model absolutely lifelike behavior when something much simpler would be more engaging for the players.
- Don’t create one complicated behavior set, when you could easily create two simpler entities. (Like what Craig had told me)
A lot of this has to do with the type of game being created, but they both encourage to go with a simple AI behavior over a complex one.
- Easily understood or grasped by intuition
James Portnow discusses how intuitable AI works and the difference it has to a learnable AI, a Learnable AI isn’t a AI that learns in this context, it’s more about the player learning about the AI, Learnable AI are basic and should allow the player to be able to learn how they move or what frame they move at depending on how much time the player spends on the game.
A Intuitable AI, is different it has a decision tree that is responsive to the player. The NPC will make a decision based on the situation at hand, in a FPS, if a player throws a grenade that has a blast radius and the NPC is in that blast radius it will go through it’s decisions and choose the best one fitting for the situation in this case, jumping to cover. Therefor it is not predictable like a learnable AI, but it also needs to be intuitable by the player, the player should be able to understand what the NPC is doing and react to the NPCs actions.
Conclusion to this, the purpose was to answer the question of which path my AI should head towards. A simple AI is the obvious choice, but I also will add some sort of complexity to it. Like Kimberly Voll suggests, start simple and then go complex. If it gets too complex, I can split the behavior into two entities to make it more simple, but I will also need to make the demo intuitable to the player, an AI should work with the core purpose of the game, to create that intuitable feeling.