This post is an interesting thing on many levels. Aside from the mind-boggling number of responses (fifty and counting), the first paragraph is a critical thing for game designers, movie writers, and foolish artists to remember:
"Toys are fun. The other side of that colorful plastic sword is when they break. Or when Santa, or the powers that be, just drop the ball and F you in your little A in the process." -Steven of The Sneeze
I think everyone has experienced this first hand, and it's always a bit strange when someone doesn't react to losing a toy. The total lack of personal investment is either creepy or disappointing, depending on the situation.
But this isn't a standard "there is no light without shadow" tidbit (although that, of course, is well worth thinking of). This is an "investment is measured by addiction" tidbit.
Your goal, in your entertainment product, is not to entertain. No. It's to fascinate. There's a subtle difference. Nobody I want to know was entertained by Schindler's List, but it was still a very good movie. Why? Because it's fascinating. You watch it with a kind of dark sense of wonder, a kind of awe at the things people can do while still looking lucid.
Once you grok this, you can get on with yourself. You're not here to make everything nicey-nice. You're not here to provide humor, or cool space fights. Those are tools. That's like saying a computer is Microsoft Word.
You're here to fascinate. If that means darkness, it means darkness. If it means humor, it means humor. If it means cool fight scenes, put them in. If it means a character needs to die, then by all means, kill them. You need to get the audience totally involved. Only then will their investment be complete. Every further step you take it will be further investment. The most glorious toys are the ones lying in fragments in your old toy box.
People remember FFVII because Aerith died. Otherwise, it wouldn't have the same kind of following. They thought she would come back. She didn't. It was a great move. It created a kind of horrified fascination with the game. "What is this? This isn't how games go! What's going on?"
It's the same with Silence of the Lambs. It was a fascinating dip into the darkness of abnormal psyches. It was not about cool fight scenes, although there was one.
Your need, when you want to entertain, is to keep the audience fascinated.
Patterns can be built and challenged and twisted on every level, whether gameplay or character interactions. But the key is: don't be afraid to commit. Undercommitting is a crime. It shows a weakness, a lack of authority, a lack of vision. It shows you're afraid of the audience. It shows that you have nothing to show them.
Overcommitment can be bad, it can be irritating... but it will always be fascinating. And even irritating media can still keep an audience, if it is fascinating. Undercommitting is in no way fascinating. It's predictable, it's washed out, it's weak.
This means you should take your patterns and change them permanently, fundamentally. Characters can die. It is not against the rules. Villains can win. It's not against the rules. You're allowed to put in true reversals, horrible jokes, nudity, and true madness. It's not against the rules. Don't be afraid to push your patterns too far.
Because "too far" is always worth watching.
Qualifier: a lot of beginners take patterns "too far", but they take them "too far" in an extremely boring way. You still have to follow the basic rules: patterns are built to interact with other patterns. If there are no meaningful interactions to make because you've placed your pattern too high above the rest, then there is no meaningful reason to watch the media.
Of course, you don't have to have the interacting patterns inside the media. They could be patterns in the audience's head. That's a powerful technique, but one to be covered at another time.
Qualifier the Second: You don't need to (and probably shouldn't) always "see-saw" your patterns. "There is no light without shadow" is a good thing to remember, but it's largely to set up contrast and define a pattern's power level and methods of interaction. Not everything has to be "light-dark-light-dark". Your hero does not need to have a failure for every success, or a success for every failure.
All you need to do is offer consistently unique and fascinating pattern interactions.
In terms of Pattern Adaptation Control, the most efficient method of offering new and unique interactions is to regularly make patterns shift and change. This gives you a "new" pattern to use without losing the emotional investment the audience has in the old pattern while SIMULTANEOUSLY getting the player to invest in the "pattern of changes in this pattern" pattern.
It's extremely efficient.
Qualifier the Third: Damn, this is a lot of qualifiers.
If your audience loves someone, and you screw them, there will be a giant backlash. The key is to keep that backlash in the magic circle. You want zero negative emotions aimed at you, the author. Positive emotions are okay to leak everywhere, but negative emotions need to be carefully coralled into leak-proof containers.
The 'walls' of these containers are made out of a handy receptical, usually a "villain". However, anger still spills out of the top. Your villain has to be "big" enough to hold all the anger and hate. You have to stretch him beforehand so he can take it without overflowing and spilling on you. If you're killing a popular character, the villain had better be a known badass.
It's also often a good idea to put a "funnel" on to prevent any splashing during the actual event. The most common kind of funnel is an in-game reference. Aerith dies, and all the characters go, "No! Aerith! I hate you, Seph! I'll kill you!" By showing other characters feeling the same thing the player is feeling, you can lead players to put their emotions in the same container as the characters do through basic sympathy and resonance.
You can also milk this later. You've established a conduit of hate, and by tugging on it at any point, you can add some more hate to your villain, who is by now plenty stretchy enough to handle the additional ire. This can also enhance your good guy, which is another method that will have to be covered some other time.
This methodology can (and should) also be used for good emotions.
It's also a chapter on its own - this is just a quick overview. :)