I've called this second property "answer diversity." Now let's sort by it, too: Doing this, we can think of our space of questions as four zones, roughly described like so: Clearly, the lower right-hand corner contains the kind of questions we want, and that's where we found the correlations we report below.
The basic currency of the Internet is human ignorance, and, frankly, our database holds a strong cash position!
So, instead of judging each question's first-date appropriateness subjectively, I turned to statistics.
I decided our candidates were the ones that (a) most people were comfortable discussing publicly, and (b) were mathematically likely to tell you something you couldn't just guess.
There is so much you want to know about the person across the table from you, and yet so little you can directly ask. We took Ok Cupid's database of 275,294 match questions—probably the biggest collection of relationship concerns on earth—and the 776 Love, sex, a soulmate, an argument, whatever you're looking for, we'll show you the polite questions to find it.
We hope they'll be useful to you in the real world.
Before we could go looking for correlations to deeper stuff, our first task was to decide which questions were even first-date appropriate.
I know each person has his own opinion on what's okay to talk about with a stranger.
I sliced Ok Cupid's question pool like this: That blue rectangle is our highest-quality, least-invasive questions, and we next examined each of them for interesting correlations.
(If you're interested in knowing more about the above graph, you can drop-down an explanation here, complete with an interactive scatter plot that took me forever to make.) Whenever a user answers a match question on Ok Cupid, he has the option of keeping his answer private by clicking this box: The less often people check that box for a given question, the more confident we can be that the question is okay to talk about. Here's a simulation of the process—I can only picture a small subset of the data we crunched, but it should illustrate the principle involved.