I have an urn with 30 red balls and 30 green balls. You can pick a color, and you get 10$ if you guessed the next ball I pull out of the urn correctly. Alternatively, I have another urn with an unknown amount of red and green balls in them. You could also choose to use this urn with a color of your choice, if you want?
Which of these 4 options do you pick?
It turns out that people overwhelmingly prefer the first urn. They have an aversion to epistemic uncertainty, even though from utility theory, all 4 options are equivalent. Even weirder, if we would have played this game repeatedly, the second urn is clearly preferable. Then why do we have this intuition?
The experiment from his paper is slightly different and (in my opinion) harder to understand, but the sketch above illustrates the same paradox between what utility theory tells us is the best decision, and what we intuitively decide.
Do note that he was working at the RAND corporation at the time, where they were running probabilistic simulations of the cold war, the so called Cold War games. (People were literally throwing dice all day to run the simulations of the various nuclear war scenarios). His paradox was a critique to that method, as we don't actually know the probabilities involved in these nuclear scenarios, in the same way as we don't know the amount of red and green balls in one of the urns.
Therefore, we might want to discredit decisions based on scenarios where we don't have good estimates of the probabilities of the outcomes, in favour of scenarios where we do know them.