Open Thread 356

...

This is the weekly visible open thread. Post about anything you want, ask random questions, whatever. ACX has an unofficial subreddit, Discord, and bulletin board, and in-person meetups around the world. 95% of content is free, but for the remaining 5% you can subscribe here. Also:

1: Comments of the week, all on the Rise Of Christianity review: I originally said I was embarrassed to learn that early Christians opposed abortion, because I’d bought the liberal story that this was an artifact of 1970s Moral Majority politics. But Stephen Saperstein Frug says I was misunderstanding the story - Catholics have opposed abortion since forever, but Protestants didn’t care until the political realignments of the 1970s. And Ty Harding corrects my misunderstanding of the concubinage issue - Pope Callixtus didn’t try to sneak polygamy into Christianity, only to legitimize certain “lesser” types of monogamous marriage. And David Roman backs the role of women in early Christianity.

2: I started my discussion of the Early Christian strategy with the story of the TIT-FOR-TAT bot. But G2F4E6E7E8 on the subreddit says that the science of game theory has moved on; TIT-FOR-TAT was defeated in certain evolution-like noisy prisoner dilemmas by a strategy called WIN-STAY LOSE-SHIFT:

Why does Win-Stay, Lose-Shift win? In the simulations, it seems that at first, Tit-for-Tat establishes dominance just as the old story would lead you to expect. However, in a Tit-for-Tat world, generous Tit-for-Tat does better and eventually outcompetes. The agents slowly become more and more generous until a threshold is reached where defecting strategies outcompete them. Cooperation collapses and the cycle repeats over and over. It's eerily similar to the good times, weak men meme.

What Win-Stay, Lose-Shift does is break the cycle. The key point is that Win-Stay, Lose-Shift is willing to exploit overly cooperative agents---(defect, cooperate) counts as a win after all! It therefore never allows the full cooperation step that inevitably collapses into defection. Indeed, once Win-Stay, Lose-Shift cooperation is established, it is stable long-term. One technical caveat is that pure Win-Stay, Lose-Shift isn't exactly what wins since depending on the exact relative payoffs, this can be outcompeted by pure defect. Instead, the dominant strategy is a version called prudent Win-Stay, Lose-Shift where (defect, defect) leads to a small chance of playing defect. The exact chance depends on the exact payoffs.

Commenter bibliophile785 adds that this is also obsolete, and the very newest results are even more complicated - read the thread for more, and thanks for the correction / interesting information!