There's very little which can be directly observed. We can't wiggle our heads to gain perspective (even waiting six months to utilise orbital parallax only works for some near stars). We're working with spectral lines, redshifts, "stellar candles" (Cepheid variables, Type 1a supernovae), orbital mechanics, gravitational lensing, and a grab-bag of other tricks to sort out what's happening tens, hundreds, thousands, millions, or billions of light years away. The rest is then extrapolated based on physical models: rates of stellar fusion, time to consume fuel, behaviour of accretion disks, etc.

TFA is pretty interesting in that it notes the specific estimation methodology used to assess SMBH masses.

The fact that any of this is possible is almost magic.

Except, of course, that it's actually science.