I am reading How We Decide by Jonah Lehrer. I need to read further to decide whether it will be worth recommending, but chapter 3 has some great insights for reward systems and game design. Chapter 3 is about errors, how our simian brains will vainly search for patterns in randomness and provide excessive neurochemicals to keep us focused on what is actually unpredictable.
We are a pattern-seeking species, with self-programming neurons that seek to predict risks and rewards. We are very good at developing intuitive understandings of situations in which we do not have enough time for a rational analysis, kind of like how your dog can catch a frisbee even if he cannot do the math of plotting three-dimensional vectors in real time. Unfortunately, we apply the same mental programming to completely random sequences, seeing patterns that do not exist and feeling bigger highs from wins strictly because they are unpredictable.
I say that again: unpredictable wins produce greater emotional reactions. Your brain is programmed to look for surprises, and if it really is a random system, success will always be surprising to some degree. This is a real emotional high; you really are getting that dose of neurochemicals; your neurons really do react more strongly to a 5% chance than to filling 5% of a bar. You may prefer a badge-based raid system that lets you reliably earn rewards, but your brain will compel you to pull the lever more often for the random chance. (Developers need not know neuroscience to exploit this; see again your dog and the frisbee.) Slot machines are negative-sum games, where you will always lose money over time and eventually face gambler’s ruin, but they are the clearest example of this randomness-rewarding neurochemistry.
Meanwhile, because the results are unpredictable, any patterns you try to perceive and follow will be useless at best. If something happens 60% of the time, the best you can do is to bet on its always happening. You will be right 60% of the time. If you bet against its happening 40% of the time, it is still random which 60% it is, so you are right (.6*.6+.4*.4=) 52% of the time. This was borne out experimentally, in which human test subjects really did win 52% of the time and did worse than the lab rat that won 60% of the time. Against randomness, you have two winning bets: if it is positive sum, bet as often as you can, as quickly as you can; if it is negative sum, stop betting. Kill as many goblins and raid bosses as you can; never play a slot machine. Thoughts beyond figuring out your expected payoff are entirely wasted.
Now that you know this, have I inured your brain to make the random rewards less unexpected, thus reducing your neurochemical addiction to loot piñatas?