On the Rationalist Solution to Gregory Kavka's Toxin Puzzle

Gregory Kavka's 'Toxin Puzzle' suggests that I cannot intend to perform a counter-preferential action A even if I have a strong self-interested reason to form this intention. The 'Rationalist Solution,' however, suggests that I can form this intention. For even though it is counter-preferential, A-ing is actually rational given that the intention behind it is rational. Two arguments are offered for this proposition that the rationality of the intention to A transfers to A-ing itself: the 'Self-Promise Argument' and David Gauthier's 'Rational Self-Interest Argument.' But both arguments – and therefore the Rationalist Solution – fail. The Self-Promise Argument fails because my intention to A does not constitute a promise to myself that I am obligated to honor. And Gauthier's Rational Self-Interest Argument fails to rule out the possibility of rational irrationality.