Analysis of transistion costs for changing habits — LessWrong