It's clear that we fundamentally disagree on a huge number of issues, and that's fine, but I'll answer that one:

why would we guess or assume who people are and then interview them?

Because we are making assumptions all the time, but might not know that we're making assumptions, so it's simply to surface those assumptions and decide upfront which of these assumptions we have huge amounts of evidence to support, and which, when we think about it, we know we don't really know anything about.

Cowan suggests using BACKGROUND/THINKS/SEES/FEELS/DOES as the dimensions but I've moved towards FACTS/BEHAVIOURS/WANTS & NEEDS/FEARS & FRUSTRATIONS/DRIVERS - either captures the essence of what we want to find out.

A workshop activity I like to do is then to divide those assumptions into ones that will make a fundamental difference to the area that we're planning to explore, and which ones are pretty much irrelevant.

The riskiest assumptions - things that will make a big difference and which we don't know much about - then form the focus of the interview script.

The fun part afterwards is then to compare the interview findings with the protopersona - I highlight each sentence in green if it's substantiated, yellow if it's ambiguous, and red if it's flat-out wrong.

I've never seen a protopersona emerge with less than half of it in red.