All simulators
Counterfactuals & Attribution
Abduction, action and prediction for one worker's wage, and bounds on how many people a vaccine actually protects.
Day 59: Counterfactuals & the Ladder of Causationobserved log wage
2.610
counterfactual with 16 years
2.930
gain 37.7%
PROBABILITY OF NECESSITY AND SUFFICIENCY (vaccine trial; y = not infected)
lower bound
0.080
max(0, P(y₁) − P(y₀))
upper bound
0.100
min(P(y₁), 1 − P(y₀))
PN if monotone
0.082
share of the uninfected vaccinated protected by it
max share harmed
0.020
Left: abduction, action, prediction in Y := 1.5 + β·S + U. The worker's U is read off her observed point (abduction); then schooling is set to a new value (action) and her own line, parallel to the population line, gives the counterfactual (prediction). Right: from an experiment alone, the share of people who respond (PNS) is only bounded; monotonicity pins it down.