Re:A better way to divide the pie
by huynhhacnguyen on Sunday March 20, 2022 @10:32PM
(#30749429)
Wow thanks for the very detailed explanation, mate, that is much appreciated. However, I still wish the author had delved into more real world examples, to let us have more sense of how this principle can be applied into more practical use cases. "Algorithms To Live By" https://www.goodreads.com/book/show/25666050-algorithms-to-l... did a really good job at that. Of course an article cannot be at book length but a man can hope.
Another kind of situation is if there are 3 people, Alice, Bob, and Charlie, with fallbacks x, y, z respectively. Suppose they all need to reach an agreement to get n slices. Then g=n-(x+y+z) and Alice, Bob, and Charlie should get x+g/3, y+g/3, z+g/3 respectively.
Another possible situation is if there are 3 people, but they have unequal roles. Alice only needs to make a deal with at least one of Bob or Charlie, to get n slices. If Bob and Charlie both agree to the deal, then the total is still n slices. Intuitively, this would make Bob and Charlie less important, so they should get less of the gain. The split here should be x+2g/3 for Alice, y+g/6 for Bob and z+g/6 for Charlie, if they're all in on the deal. The reasoning there is that if we're adding games together, then we should also add the payoffs together. We can make a Alice&(Bob|Charlie) game by adding together Alice&Bob + Alice&Charlie - Alice&Bob&Charlie. Similarly, the payoffs should add like this: (g/2, g/2, 0) + (g/2, 0, g/2) - (g/3, g/3, g/3) = (2g/3, g/6, g/6)