I am also a big fun of using probabilities rather than correlations, whenever it is possible. Correlation, but itself does have an easy, actionable meaning (two variables are correlated... how much is much? what does it mean that correlation is only (or as much as?) 0.5?).
See also this ratio, P(A n B)/[P(A)P(B)]:
- http://stats.stackexchange.com/questions/6047/does-this-quan...
I used it for some projects, in which correlation resulted in a mess:
- http://stared.github.io/tagoverflow/
- https://github.com/stared/wizualizacja-wolnych-lektur