If the interviewee gets stuck just have them write a program and spit out the values averaged over many runs. This is what convinces a lot of people about the Monty Hall problem too