How do big companies, like Facebook, model individuals and their interaction?

As a layman in AI, I want to get an idea of how big data players, like Facebook, model individuals (of which they have so many data).

There are two scenarios I can imagine:

1. Neural networks build clusters of individuals by pure and "unconscious" big data analysis (not knowing, trying to understand and naming the clusters and "feature neurons" on intermediate levels of the network) with the only aim to predict some decisions of the members of these clusters with highest possible accuracy.

2. Letting humans analyze the clusters and neurons (trying to understand what they mean) they give names to them and possibly add human-defined "fields" (like "is an honest person") if these were not found automatically, and whose values are then calculated from big data.

The second case would result in a specific psychological model of individuals with lots of "human-understandable" dimensions.

In case there is such a model, I would be very interested to know as much about it as possible.

What can be said about this:

1. Is there most probably such a model (that is kept as a secret e.g. by Facebook)?

2. Has somehow tried to guess how it may look like?

3. Are there leaked parts of the model?

My aim is to know and understand by which categories Facebook (as an example) classifies its users.