Why MMO producers should invest in AI (3 - Immersion)

"It's just a game" -- we often hear that, at times spelled by own reason, when "just a game" carries forth substantially more emotion and engagement than we would have expected of it. In fact, this may just be the right effect of a good game. Games give us entertainment, learning, experience, they give us a chance to play all those scenarios that real life would have otherwise deprived us of - either because they are hypothetical, too dangerous, or outright impossible.

But they are finite, in all possible ways. Their story either ends altogether after a few weeks of playing, or becomes repetitive -- perhaps no less entertaining to some, yet still unsurprising. Their ability to grasp one's attention, personality, consciousness -- is equally finite. You rarely make decisions as YOU would make decisions. Every game player quickly learns to role play their character according to a few very restrictive rules, in a world with very few degrees of freedom.

Then there are the sandbox games, where every player can do anything, inside a rather wide rule set. They can build structures, trade freely with other players, kill them at their whim or join them in ad-hoc parties. Naturally though, the absence or rules is just about as unsurprising as the abuse. Without an obvious, enforced goal, the game is quick to become boring. And without a discrete guide to every decision, it won't be long until a player will build a medieval fort in the shape of a star ship, or a star ship in the shape of a duck or what not.

What if we could have a game engine that allows the very rules evolve and change? What if, when cleared of monsters, a forest would not re-spawn them five minutes later, but instead would brighten up and be reclaimed by the fairies who once inhabited it, but not just for one instance of a quest -- permanently. What if, instead of fighting a giant frog, you could kiss it and it would turn into a fat lady? There are many questions here: how do you implement such open mechanics? Who fills in all the possible actions? How can you script such a diverse scenario? The answer is: you don't have to!

So why should you invest in AI?

You don't even have to invest that much, because the AI can learn -- and it can do so much faster than any scenario writer. 

If you just want a challenging game, PvP will do the trick -- all you need is a barely adequate matchmaking engine that doesn't pit mice against behemoths. That will last a month on average, until players hit their best possible and become bored. PvE that just spawns enemies in different colors, at different positions is likely to last just as much. You can beg for more attention by providing a lot of mini-games, like cooking, fishing, gambling, dancing or toad racing.

There are two ways that need to work together to provide a spectacularly better player retention: the first is the change of strategy, the second is the game responsiveness. Together, they create immersion. The players' feeling that the game changes as a result of their actions and that they, in turn, need to change their approach to keep up, will make them responsible for their character and interested to see what's next. Doing this by constantly writing new, game-changing content is painful, takes a huge investment and needs precious time to implement. How many of the initial WoW players made it through to the 3rd expansion?

On the other hand, automating a minimum game change (now we play more paper than rock) will not challenge anyone. 

You need a good enough AI that can change strategy at an equal mind set with a human player.

How do you get that? Well, it comes from two directions. The first direction is the AI itself. The game needs to be represented by an ontology, that gives the AI information about ALL the degrees of freedom: what can change, what is an acceptable rate of change, what are the possible values, what are the means of change, what are the triggers, what are the consequences of each change, what is the scope of each change? Think of a game world in which one night, unbeknownst to any developer, the AI decides to start a massive migration of tribes from one side of the world to another, causing a mayhem of raids, clashes and all out battles, while also modifying the landscape substantially. How can this be done? Just define "relocation" as a possible change. Poverty or famine can be a trigger, together with the knowledge about a better place; the scope can be a single AI character, a family, a tribe, an entire civilization, by simple majority vote, or weighted vote according to social standing. 

This allows the developers to "cause" change by manipulating the ontology, instead of implementing it explicitly -- an indirect and far more subtle touch on the game world, one that computes the player performance as well, not just the 

The second direction is the game structure. You need to construct the game in a loose enough manner that change can happen, but linear enough at the same time that, when it happens, change will compel all players to react to it. Think of the game as a wind tunnel: you can move around freely, but you cannot dodge the wind. The game needs to give the AI a chance to asses its own performance against human players and make changes to all the ontology elements that it can change, to optimize for the next round. The "round" concept entails a certain segmentation. It can be seamless, but it's better to be declarative -- telling the players that the game change is a part of the game. 

Here's a brief comparison of a classical MMO versus an AI driven MMO:

Classical MMO:

• Game change performed by writing new content
• Same rules, same challenges, different colors
• No surprise and no game changing effect from the player
• Limited replay value

AI driven MMO:

• Game change performed automatically by the AI
• Requires different approaches, different strategies from the players
• The players change the way the world works with their actions
• Every round is different, for as long as the AI has the liberty to change the world.

The AI using the ontology to change the game, combined with the segmented narrative (the rounds) -- create the immersion effect, the illusion that the persistent world of the game evolves, is not repetitive, so the same approach will not work a second time and that the actions of each player affect the world irreversibly. The immersion effect, in this case, is equivalent to manipulating the players' empathy, by linking them to the characters they are controlling and creating a sense of responsibility towards a world that reacts permanently to their existence, much like, in a surprising way, the real world.

So why should you invest in AI?

Because it can help you create a new generation of games, that evolve on their own.

Because it can keep players in the game much longer, at a much smaller development cost.

Because there is makes content writing obsolete, in favor of ontology writing, which is reusable.

Because instead of writing rules for static content, you will start writing rules for diversity and change.

Because it is adaptable to real life games, such as military simulations, or stock exchange strategy.