How has machine learning and AI influenced game design? originally appeared on Quora: the place to gain and share knowledge, empowering people to learn from others and better understand the world.
Answer by Travis Addair, Software Engineer, on Quora:
Games are more addictive today than ever, and we have machine learning to thank for a lot of that.
When most people put “AI” and “games” together they tend to think “bots” or some other form of computer-controlled character (NPC).
But the reality is that machine learning is virtually unused for such “AIs”. As I’ve discussed in other answers [1], it’s just too expensive and doesn’t allow for the degree of control required by the design team.
However, that’s not to say that machine learning is entirely absent from the game development process. In fact, it’s used quite extensively, just not in the game itself, but in the way we design gameplay systems.
Let’s call it data-driven game design.
See, game companies — particularly the big ones like Activision and EA — employ many data scientists.
What do all these data scientists do for the company? They don’t directly make games, rather their role is to inform the game design by understanding player behavior.
How do they do that? Today, more and more games are moving their systems to the “backend” — the server side — where all the information about what the player does in the game and when they choose to play or not player the game can be tracked, recorded, and analyzed.
That analysis can be simple — like a linear model used to correlate variables — but increasingly it’s becoming very complex. As more data is tracked in greater quantities, more data science teams are turning to machine learning models like gradient boosted trees and deep neural networks.
To what end? Keeping players engaged with the games, keeping them coming back for more.
This can be done at a macro level (looking at averages, or segmentations of users) or at a micro level (individuals).
What if you could predict when a player was about to stop playing your game? You could entice them to come back with custom bonus offer or in-game reward.
What if you knew just the right frequency of rewards needed to keep a player interested and coming back each day? You could customize loot drops to suit the preferences of specific users.
For better or worse, that’s the direction the industry is going, and we have machine learning to thank for it.
Footnotes
[1] Travis Addair's answer to Are there any video games that use neural networks for their AI?
This question originally appeared on Quora - the place to gain and share knowledge, empowering people to learn from others and better understand the world. You can follow Quora on Twitter, Facebook, and Google+. More questions:
