ATC is auto temp correction. (edit - it works because there is a bimetallic strip, like what's in a thermostat, that changes length/curvature depending on how warm/cold it is...this moves a lens that's focusing light from the prism, changing where the "blue line" falls on the scale)
bpolivka is exactly right. This corrects the reading for changes in temperature for the refractometer. See, the reading from a refractometer depends on the refractive index of both the sample and the thing the sample rests on, (the prism). ATC corrects for the fact that the RI of the prism changes with temperature.
We can put hot wort on a refractometer because the amount we put on is very small. It cools quickly, and everything is OK.
PROBLEMS:
If you dump a whole bunch of hot wort on the refractometer, like if you put 20 drops on it, you'll heat it up too much and get funny readings.
If you take multiple readings within a short period, you'll heat it up and get funny readings, (I try to rinse the prism with room temp water between samples to get it back to a good temp).
Likewise, washing it with hot water, then taking a reading, will screw up your reading.
(noticing a trend?)
Keep the prism and refractometer body at a decent temp, (within 10 or 20 deg of room temp is a good rule of thumb, but there are usually designated ranges that ATC can work in). Use the minimum amount of fluid required to take a sample, and wait a few seconds at least, (I wait 30 to 60 seconds at least) between placing the sample on the prism and taking a reading.
By the way, evaporation is REALLY minimal. When you have that coverplate over top of the prism, only the edges are exposed to air and able to evaporate. So you have minimal surface area exposed for evaporation. Add in the fact that you'd have to have the whole width of the sample change SG to get a different reading, and that you're now mass transfer limited for water to move to the edge of the coverplate to evaporate, and it's a non-starter.