It looks like you are trying to apply Bayes' theorem which tells us that
$\displaystyle P(A|C) = \frac {P(C|A)P(A)}{P(C)}$
where event A is "family has two daughters given that the family has two children and at least one daughter" and event C is "a daughter answers the door given that the family has two children and at least one daughter".
You know that $P(C|A)=1$ and $P(A) = \frac 1 3$. If $\lnot A$ (which I think you are calling B) then the family has one son and one daughter, so $P(C|\lnot A) = \frac 1 2$ (this is where we use the information that one of the children will answer the door, and each child is equally likely to answer the door). So we can find $P(C)$ as follows
$\displaystyle P(C) = P(C|A)P(A) + P(C|\lnot A)P(\lnot A) = \frac 1 3 + \frac 1 3 = \frac 2 3$
which gives us
$\displaystyle P(A|C) = \frac 1 3 \times \frac 3 2 = \frac 1 2$
Another way to see this is as follows. There are two scenarios in which a daughter answers the door:
- The family has two daughters and one of them answers the door - probability $\frac 1 3$.
- The family has one son and one daughter and the daughter answers the door - probability $\frac 2 3 \times \frac 1 2 = \frac 1 3$.
Since these two scenarios are equally likely, the probability that we have scenario 1 is $\frac 1 2$.