There doesn’t seem to be much of a rationality (as seen by LW) cluster. Notably, one-boxing positively correlates with many positions that would be unpopular here.
They had everyone predict what the results would be. Here’s what correlates with this predictive accuracy. The strongest effect seems to be from age/experience, which I guess is plausible in retrospect. Switching on the trolley problem is one of the positions most strongly associated with predictive accuracy (though still only .2), so maybe that’s a better place to look for a rationality cluster.