Understanding Game AI and Design Part 3: Examples and Conclusions

Intro and State Machines

Planning Systems and Terrain

Examples and Conclusions

The Splinter Cell Example

I suppose the following example will apply to most, if not all, stealth games ever made, but I guess this recent example will have to do.

Obviously, given the information above, it’s pretty obvious that stealth games operate on two different modes for its AI opponents: regular patterns, and the different ones when you, the player, are detected or make them curious enough to follow you around. All of these seem like perfectly sound video game mechanics, but I can’t help but feel the “realism” of these events never quite comes through. Even the user interface seems to admit as such, often using a “detection” meter that will, in some shape or form, tell you that you’ve been seen or detected by opposition forces.

Think about the difficulties in actually implementing such a system of realistic detection; would most soldiers immediately yell about it to their friends, or might some of them act as if nothing’s happening only to spring on you at the last second? You don’t get that sort of variety in most stealth games – either they don’t see you, they know you’re there, or you’re actively being chased in plain sight. Three different versions exist, and they look the same in most. Once they start patrolling again as if nothing happens (Metal Gear Solid, I’m looking at you), the “immersion” is lost.

This is true even of the most modern stealth games, such as Splinter Cell: Blacklist. The trailers would have you believe that a REVOLUTION IN STEALTH GAMES just happened, while it plays a whole lot like all the Splinter Cell games that came before it. Rather, it relies on the stealth genre’s tropes while adding a billion new options and upgrade systems so you, the player, make the game much more interesting under constraints both natural and artificial to the genre. The AI detects you more readily on higher difficulties, but the AI did not magically increase in intelligence; it’s an obvious artificial switch that just reduces the time you can stay out in the open.

But here’s the real question: would you PREFER a more realistic AI in a stealth game? I wonder this, precisely because the definition would vary so much. It’s obvious that Splinter Cell uses a goal-oriented planning system, but the detection still relies on an exterior element. If it were AI based, you’d see them react accordingly, but it works based on an arbitrary time limit based on your position relative to a character’s front (i.e., “viewing range”, as you would see on an MGS radar). I’ve never seen a stealth game do it differently, and Ubisoft certainly took no risk with this one. Would a realistic detection system actually make the game more fun than it was otherwise? How would you even implement this without the predictable nature of these various guards and terrorists, anyway?

In my opinion, realism should not suddenly prevent “fun” – the mastering of fair game dynamics. A realistic AI isn’t necessarily a great fit for every game; even if you could turn any old enemy or unit into a digitally-enhanced super solider, that doesn’t mean the player would automatically enjoy themselves. Gamers thrive on pattern recognition, learning, and the ability to predict future events through their current actions. Introducing that element of “perceived” randomness may improve some games while destroying others.

It all starts with the end goal of the actual game designers. You don’t ever need smart AI; you need smart design. However, a well-constructed AI can aid a game that would otherwise be rather droll, and can enhance a classic style even further. Either way, it’s a boon to players, not a detriment, in whatever form you see it. If it’s too smart, the game becomes far too frustrating and difficult; if it’s too dumb, you see through the facade far too easily. That sort of balance remains difficult to achieve.



It’s interesting that the Bible often presents two seemingly contradictory messages: ask and you shall receive, but that asking also remains dependent on God’s will and timing. Most people use that “ask, seek, knock” verse to prove that they should receive whatever they ask for, but things are slightly more complicated then that.

We know that God does not hear sinners; but if anyone is God-fearing and does His will, He hears him.

John 9:31

We often like to think of ourselves as the Lone Ranger, the one guy/gal who follows God’s will to the letter and, therefore, also deserving of answered prayer. Trouble is, God’s long-suffering and answers prayer based on His time and your attitude. Seeking a prayer request just to fulfill yourself? I doubt that’s going to happen!

You lust and do not have; so you commit murder. You are envious and cannot obtain; so you fight and quarrel. You do not have because you do not ask. You ask and do not receive, because you ask with wrong motives, so that you may spend it on your pleasures. You adulteresses, do you not know that friendship with the world is hostility toward God? Therefore whoever wishes to be a friend of the world makes himself an enemy of God. Or do you think that the Scripture speaks to no purpose: “He jealously desires the Spirit which He has made to dwell in us”?

James 4

God desires all of you, not merely a part. If you ask with wrong motives, then you will get no results; it works the same way in most relationships, after all. Ulterior motives without any good reason to have them can often block a true connection or a real relationship.

So it is with AI in video games. What exactly do you want, really? Is the AI in games somehow insufficient? We’ve demonstrated pretty handily how much the current system works, and how we don’t need a truly smart AI – we just need a well-programmed one that takes advantage of smart game design. Any and all games rest solely on the shoulders of the developer’s well-thought plan, not some magic artifical intelligence that can fix all games forever – otherwise, things would turn boring. We always desire something beyond what’s necessary, just because. That’s the problem with a unfulfilled heart – it looks for fun in all the wrong places.

Maybe we should just see what’s there instead of hoping for something that wouldn’t fix bad game design if it tried.

About Zachery Oliver

Zachery Oliver, MTS, is the lead writer for Theology Gaming, a blog focused on the integration of games and theological issues. He can be reached at viewtifulzfo at gmail dot com or on Theology Gaming’s Facebook Page.