It's really just about splitting the risk across 2 tries instead of 1.
Suppose you're a team that converts 60% of your 2 point conversion attempts.
If you go for for it when you score to go down 1, you have a 60% chance of winning (40% chance of failure)
If you go for it on 2 tries, there's an 84% probability of converting one or both (drops probability of failure to 16%)
Of course if you succeed on the first try (down 8) which has an equal probability of success as down 1, you get an even higher chance of going up because PATs are 99%+
At least that's what I think the rationale is