Perhaps the currency of social sciences should be anecdotes with full context (of which you can obviously get only a limited number unless you have unlimited budget) rather than collecting extensive data points on a limited number of variables. While physical systems can be approximated (or some independent variables dropped) without affecting the aspect being studied (think perfectly spherical objects), when it comes to humans, there are no independent variables. Approximation or simplified models are much harder.

I think looking for statistical patterns (e.g. in literature) is perfectly good science as long as you are cognizant that patterns merely invite more study an should not be used to reach conclusions, also being aware that patterns might disappear when you expand your data set.

Finally, as someone trained in the physical sciences, I used to look down on social scientists. I no longer do this. At least they're brave enough to tackle a complex monster with the limited tools at their disposal, stumbling and even enduring ridicule from the hard sciences. We ignore the human mind and collections thereof, because it's too complex and prefer the relative comfort of simple, predictable systems. I don't believe that's good.