TL;DR, I believe: They defined metrics[2], gather the relevant values from applicants and ask their algorithm: "How much?"
Buried in [1] is this gem:
In our pilot, we evaluated nearly 3,000 companies and committed to funding several dozen of those, across 12 countries and many sectors, without a single traditional venture pitch. In fact, in most cases the data-driven approach allowed us to reach conviction around an investment opportunity before we ever even spoke to the founders. Worth noting that when we recently looked at CEO demographics, we found that 42% were female and the majority nonwhite.
A purely data driven approach to VC investment could level the playing field a lot and I could imagine that adapting that concept to hiring is a) not too far off and b) a great way to reduce biases in the hiring process - if everything works as promised (or rather "as hinted at").
[1] https://medium.com/social-capital/capital-as-a-service-a-new...
[2] https://medium.com/swlh/diligence-at-social-capital-part-1-a...