This requires a hypothesis test. You will have to assume (I assume) that the 2 random variables ( and , the times that the Milton and Webster - respectively watch per day) are Normal random variables.
Let be the population mean of the time that Milton watches per day. (i.e. his true average time per day), and let be the same for Webster.
In theory you could do this using matched pairs (1 sample test based on the differences since there is a nice pairing (assuming that the father observed his kids on the same 21 days. However, the problem seems to be pushing you to do a 2-sample test.
Can you finish the problem?? (Note that )