The payoffs for “other” were designed so that neither failing to halt, nor convincing the other player not to halt, should ever be a worthwhile strategy. If you don’t halt, it gives you the same payoff as if you had cooperated, and gives the other player the same payoff as if you had defected. That way, not halting should be strictly dominated by defecting, since you are better off if you defect, and the other player should react the same way to each threat. And tricking the other player into not halting is also a bad idea, since the payoff you get from it is the same as if they defected.
The payoffs for “other” were designed so that neither failing to halt, nor convincing the other player not to halt, should ever be a worthwhile strategy. If you don’t halt, it gives you the same payoff as if you had cooperated, and gives the other player the same payoff as if you had defected.
Your game world implements an “enthusiastic consent” policy
True, but I still don’t expect it will be a big problem. If there are a lot of submissions, the effect will be small, and if it is paying enough attention to your source code for it to be possible to trick it into not halting, then it is probably looking for a way to achieve mutual cooperation, so tricking it is still not a good strategy.
If you trust all sufficiently smart players to try to induce (Defect, non-halt) if (Defect, Defect) is otherwise inevitable, the effect adds up over a hopefully significant portion of the submissions.
The payoffs for “other” were designed so that neither failing to halt, nor convincing the other player not to halt, should ever be a worthwhile strategy. If you don’t halt, it gives you the same payoff as if you had cooperated, and gives the other player the same payoff as if you had defected. That way, not halting should be strictly dominated by defecting, since you are better off if you defect, and the other player should react the same way to each threat. And tricking the other player into not halting is also a bad idea, since the payoff you get from it is the same as if they defected.
Your game world implements an “enthusiastic consent” policy
(Defect, non-halt) is actually better than (Defect, Defect) for you, since it gives you a relative advantage over competitors in the tournament.
True, but I still don’t expect it will be a big problem. If there are a lot of submissions, the effect will be small, and if it is paying enough attention to your source code for it to be possible to trick it into not halting, then it is probably looking for a way to achieve mutual cooperation, so tricking it is still not a good strategy.
If you trust all sufficiently smart players to try to induce (Defect, non-halt) if (Defect, Defect) is otherwise inevitable, the effect adds up over a hopefully significant portion of the submissions.