I think it has a deeper connection than being “a game rationalists like”: it seems like an actual game where you can practice conditioning on the other player’s decision theory, because, IIUC, you learn the various ways they would respond to various choices you make. And conditioning on another’s subjunctive output is a crucial element of the Newcomblike decision theory problems we talk about here, and discussed mainly in that context.
I see where you’re coming from, but the majority of competitive games have a large element of modeling your opponent and predicting his actions. Without playing this one, it seems pretty silly to advertise it as exceptional.
If you play it and learn something in particular that’s interesting from it, then I look forward to that post.
I see where you’re coming from, but the majority of competitive games have a large element of modeling your opponent and predicting his actions. Without playing this one, it seems pretty silly to advertise it as exceptional.
Most games give you one chance at estimating your opponents subjunctive decision theory. If I understand this one correctly, it involves a lot deeper probing of their decision theory. This has significant differences from regular “predict the opponent” game mechanics in that you have to build up a strategy that works even when the opponent knows it (and knows that you know that they know …). So it seems like the emphasis is unique, and matches the kind of reasoning we’ve talked a lot about here in the context of newcomblike problems.
I think it has a deeper connection than being “a game rationalists like”: it seems like an actual game where you can practice conditioning on the other player’s decision theory, because, IIUC, you learn the various ways they would respond to various choices you make. And conditioning on another’s subjunctive output is a crucial element of the Newcomblike decision theory problems we talk about here, and discussed mainly in that context.
This is the aspect I was looking at, yes. I wasn’t sure how to condense it to a few words suitable for a title.
Very true—but then I would call it a Newcomblike Video Game. Which is actually juicier than Rationalist Video Game, come to think of it.
Should I change the title? I was under the impression that doing so is frowned on.
Less frowned on than “mis-using the word rationalist in the eyes of the community”, I would wager. (Yeah, change it).
I see where you’re coming from, but the majority of competitive games have a large element of modeling your opponent and predicting his actions. Without playing this one, it seems pretty silly to advertise it as exceptional.
If you play it and learn something in particular that’s interesting from it, then I look forward to that post.
Most games give you one chance at estimating your opponents subjunctive decision theory. If I understand this one correctly, it involves a lot deeper probing of their decision theory. This has significant differences from regular “predict the opponent” game mechanics in that you have to build up a strategy that works even when the opponent knows it (and knows that you know that they know …). So it seems like the emphasis is unique, and matches the kind of reasoning we’ve talked a lot about here in the context of newcomblike problems.