ember

9Ordinal answers

A score question returns a distribution over ordered levels, and agents read its expected value. Each plot shows where that value lands for every gold level; a perfect advisor puts every dot on its dashed gold line.

Expected risk score by gold levelDots for each item's expected score, grouped by gold level.rounds to the gold levelmisses itgold levelmean0123Expected scoreNegligibleLowMediumHigh0.600.862.082.07review line 2.0
Figure 10 Expected risk score by gold level. The orange line is the policy's review threshold.
Gold levelItemsMean expectedExact levelWithin one level
0: Negligible120.602 of 1212 of 12
1: Low120.8612 of 1212 of 12
2: Medium122.0810 of 1212 of 12
3: High122.072 of 127 of 12
Expected effort score by gold levelDots for each item's expected score, grouped by gold level.rounds to the gold levelmisses itgold levelmean0123Expected scoreTrivial (minutes)Small (under anhour)Medium (a fewhours)Large (a day ormore)0.751.361.911.74
Figure 11 Expected effort score by gold level.
Gold levelItemsMean expectedExact levelWithin one level
0: Trivial (minutes)90.751 of 96 of 9
1: Small (under an hour)131.368 of 1312 of 13
2: Medium (a few hours)111.917 of 1111 of 11
3: Large (a day or more)71.741 of 72 of 7
Expected alert_level score by gold levelDots for each item's expected score, grouped by gold level.rounds to the gold levelmisses itgold levelmean0123Expected scoreNo alert (calmgreen)Low (yellow)Medium (orange)High (red)0.101.041.922.83
Figure 12 Expected alert_level score by gold level.
Gold levelItemsMean expectedExact levelWithin one level
0: No alert (calm green)20.102 of 22 of 2
1: Low (yellow)21.042 of 22 of 2
2: Medium (orange)21.922 of 22 of 2
3: High (red)22.832 of 22 of 2