It's a flat fee. That is, you're paying the same whether you fetch a single work or 10,000. Aaron Swartz's mass document mining would be supported. If you're interested in large-scale literature analysis, it's possible.

It's scaled to income / wealth. Information is part of what makes the system in which you're living possible. If the system's rewarded you, you reward the system back. Hence the tax element. Possibly scaling to some sort of connectivity plan might accomplish a similar aspect via price discrimination (of the old-school multi-tier service offering model, not the new-school hit-them-with-hidden-fees approach).

There are tiers of compensation. Pumping out white noise / easy-to-generate content doesn't necessarily buy you high compensation. Detailed scientific / journalistic work does. Penalties for inaccuracy in genres where accuracy matters, especially against deliberate distortion.

Implementation is the tricky bit.

Ex: I'll agree to pay $X/month, and then at the end of the month, that money gets paid out to the artists I listened to based on how often I listened to them.

It probably wouldn't work, but I don't really feel any current method of music sales is really working for most artists.