This is a neat question, but I think programs being successful is not really about gracefully going down a hierarchy. For example, Tit-for-Tat does not take the correct strategy against always-cooperate (If your opponent is always cooperating, you say thank you and always defect). Tit-for-Tat succeeds for much more ecological reasons. I’d say bigger-memory versions of Tit-for-Tat are going to be something like the class of “peaceful, non-exploitable” strategies. Such strategies are not going to be the first to defect, which means they actually don’t get that much information about their opponent. I think the lesson of iterated prisoners dilemmas is that you don’t need that information anyhow, as long as your strategy occupies a good ecological niche.
This is a neat question, but I think programs being successful is not really about gracefully going down a hierarchy. For example, Tit-for-Tat does not take the correct strategy against always-cooperate (If your opponent is always cooperating, you say thank you and always defect). Tit-for-Tat succeeds for much more ecological reasons. I’d say bigger-memory versions of Tit-for-Tat are going to be something like the class of “peaceful, non-exploitable” strategies. Such strategies are not going to be the first to defect, which means they actually don’t get that much information about their opponent. I think the lesson of iterated prisoners dilemmas is that you don’t need that information anyhow, as long as your strategy occupies a good ecological niche.