The circuit in your image won't work. Diode forward voltage for one, and if you get that working there's still a second resistor in parallel to your pot. Under DC conditions, it would only work if (a) you were keeping lower branch of pot above diode forward voltage and (b) DC signal is constrained such that the current through the LED branch is within tolerance.
LEDs do dim in response to less current, but you get minimal control. Attempting time-variant control in this way may damage the diode in some cases, most likely by hastening the decay of its luminescence. Importance of this depends on application and LED color (red has a long decay, white much shorter).
However, the better reason to use PWM is that you should be able to simply compare the voltage on the two legs of the pot via the ATTiny. This solves both your sensing and your output problems. You can probably also do nifty lin-log math and calibration so that the apparent volume maps comfortably to the shift in brightness.
For analysis, think about steady-state signals. What happens if a fixed DC voltage or a fixed AC voltage of fixed frequency is applied? Look at each separately, then consider that the actual effect will be the superposition of the two. This can get complicated, but for simple circuits you can often get a feel for it at a glance.
