The meta-game of game design: how do you design a game whose mechanics lead to incentives that align with your intentions is a fascinating area. It also has deep connections to policy design in politics.
In the political arena, it gets so much more complex because you have disagreement on intentions and people deliberately obfuscating their intentions. So you've got people arguing over legislation who are trying to aiming the laws have different emergent properties.
Right. Do you have any good recommendations to read on this matter?
Here are a couple of good papers on how metrics distort systems and what can be done to mitigate this. [1] [2].
And then a new post by Cedric Chin on practical examples of mitigating Goodhart's law in Amazon. Really good read: [3]
But generally, I feel like not many people or organizations care about this. The non-profit sports world is conservative and slow, organizations and corporations are also often slow, and there some psychological/social effects linked to the metrics-that-don't-work-anymore.
Spiritually related: examples of RL algorithms exploiting glitches in the physics engine (clipping into the ground to shoot forward at incredible speed) or gaming the goal state: (Agent that's supposed to learn how to walk instead evolving a body with a single long leg that just falls forward-- getting a high speed score with low complexity penalty) https://arxiv.org/abs/1803.03453https://www.deepmind.com/blog/specification-gaming-the-flip-...
A closely related question is: Should playing the fun way, and playing the optimal way be the same thing? I used to think think the answer was yes, but Spelunky 2 made me doubt; in Spelunky using a bomb or a rope is rarely optimal, but it can make things simpler and more fun. I think it's a well designed game, so the fact that optimal play and fun play are different is making me think.
> Should playing the fun way, and playing the optimal way be the same thing?
This is something that game designers, especially in roguelikes where there is a lot of procedural generation and combinatorial gameplay experiences, think about a lot.
There are all sorts of related questions:
* What should the space of optimal strategies look like? A single point that players should try to discover and optimize for? A region where there are a variety of equally valid ways?
* What are the discincentives for non-optimal play? Should it just be boring, or should the game actively punish the player for not following an expected strategy?
* What to do about strategies that are extremely effective but not fun? In roguelikes, that's things like "farming" where you find an easy to kill monster that breeds and just mow through hundreds of them to grind XP. Should the game try to avoid those scenarios so that players don't have to make an uncomfortable choice between maximizing versus fun, or should that be up to players?
It took me a decade to realize my best friend and I play friendly games differently. It was most obvious in magic the gathering - he would consistently optimize the same deck week after week, whereas I'd be bringing new decks with new and fairly general mechanics all the time. Similarly, he only ever played one character in smash bros or overwatch whereas I would heavily rotate.
Ultimately his goal was win percentage and optimal play - mastery of specific thing. I was going for fun and interesting wins and variety, and didn't care if I lost 90% of the time of that one win was AWESOME :-)
(Possibly relatedly - he brave a very very good Java specialist fter a decade of specific experience. I became a syaadmim then architect them ops manager type generalist :).
Absolutely! I ended up getting the "Timmy the power gamer" unglued card and put it on a landyard :-P - but in reality I'm a mix: I'm in it for fun like Timmy, but less with big creature and more with deck building and creative wins with unpopular cards that rarely pay off, like Johnny.
For games at least, attempting to be highly prescriptive in the rules risks strangling the game. Basketball was designed to be played without dribbling. By the time players started single-handed dribbling it was pretty clear what's the more interesting, exciting game to watch and play
I like the idea of a two-phase game, where for the first phase (most of the game), you're motivated to play your hardest for the entire duration, because the score imbalance wouldn't mean you win, but would just mean you have that much more of a relative advantage in the second phase.
And then the second phase could be evenly-matched, or it could be David and Goliath, but then still anything could happen.
Politics is absolutely about policy design, but there are many many higher-order social emergent phenomena on top that obscure it.
Imagine you had an organization in Washington DC called "Other Congress". It has the same giant beautiful buildings as Congress. It has the same rules and and parliamentary procedure, the same number of members, the same election schedule. In other words, it is 100% identical to normal Congress. It even passes bills. Except... those bills have no force of law. It's all make believe.
That Other Congress would never attract any significant traction or participation.
Note that this sounds sort of like a shadow cabinet, but it's not. The whole point of a shadow cabinet is to act as a social mechanism for influencing the real cabinet.
It may be that an Other Congress would end up functioning sort of like a shadow cabinet. But even then, the only reason anyone would participate in it is to affect the real Congress, and the reason they do that is because real Congressional policy carries weight.
In the political arena, it gets so much more complex because you have disagreement on intentions and people deliberately obfuscating their intentions. So you've got people arguing over legislation who are trying to aiming the laws have different emergent properties.