1
Means or side effect is a question about dependence
Either the good outcome needs the harm, or it comes about without it. The test is to intervene on the harm and re-evaluate the good, which is the but-for test applied to a harm.
Lesson · Ethics and the Arithmetic of Welfare · Deontology
Double Effect, Means and Side Effects
The doctrine of double effect allows a harm that is a side effect of a good act and forbids the same harm used as the means. The lab reads that distinction as dependence: write the case as structural equations, flip the harm, and see whether the good outcome changes. If it does the harm was the means, and if not it was a side effect.
flip the harm, recompute the good
the good flips: the harm was the means
the good stands: a side effect
intentions are not in the equations
Means or side effect, as a dependence
Core ideas
What this lesson establishes
Science, Induction and Causation gave the but-for test. The three ideas below move it from a cause of a fire to a harm that a good act brings with it.
2
The equations carry the contested work
Two cases get different verdicts only because they were written with different equations. The lab confirms that the models behave as written; whether they were written fairly is the part it cannot check.
3
Intention is not read
Double effect is stated in terms of what the agent intends, but a causal model contains no mental states. What the lab tests is the dependence that the intention is supposed to track.
The material
A harm, a good, and the equations between them
The switch and the loop as small models, and a pair of bombers.
The doctrine of double effect, usually traced to Aquinas, says an act with both a good and a bad effect can be permissible when the act itself is permissible, the bad effect is not intended as a means to the good, and the good is proportionate to the harm. This lesson takes one clause: the bad effect is not the means.
Take the switch case from “The Trolley Problem as a Decision Matrix”. Let L be “the lever is pulled”, M be “a man is on the side track”, E be “the trolley goes onto the side track”, H be “the man is hit” and F be “the five survive”. The model has three equations.
E = L
H = E and M
F = E
Everything is true in the actual case. To test whether the harm is the means, ask what happens to the five if the harm is set to false while the rest of the model is left alone. F = E, and E has not changed, so the five still survive. The lab reports that the man’s being hit is not a but-for cause of the five surviving, and not a cause at all. The five live because the trolley left their track; the man’s death is something that happens on the way.
Now the loop. The side track curves back to the main one, so the trolley would reach the five anyway unless something stops it, and the man’s body does. The first two equations are the same, but the third changes.
E = L
H = E and M
F = E and H
Definition · Means, in the lab’s sense
A harm is a means to a good when the good depends on it: set the harm to false, hold everything else as it was, and the good fails. It is a side effect when the good comes about whether or not the harm occurs.
Flip H in the loop and F becomes false: the five die. The harm is a but-for cause of the good, so it is a means. The two cases have the same lever, the same body count and the same outcome for the five. They differ in one equation. In “The Trolley Problem as a Decision Matrix” the loop was put on the forbidden list by hand. The structure says why someone who accepts the doctrine would put it there.
Example · Two bombers
A strategic bomber destroys a munitions factory, and civilians next to it die; the war is shortened because the factory is gone. A terror bomber kills the same number of civilians in order to break morale; the war is shortened because morale breaks. In the first model the war’s shortening does not depend on the civilian deaths, in the second it does. The two presets print different verdicts for the same deaths and the same benefit.
Three limits apply. The lab tests one clause. It does not test whether the act is itself permissible or the good proportionate to the harm. It also cannot say that a means is worse than a side effect, which is the doctrine’s central claim and the one most often disputed; many people judge the loop permissible, and the doctrine has to say they are wrong. And the equations are a description of the case. Redescribe the switch so that the five survive only if the man’s weight slows the trolley, and the verdict changes with no change in anyone’s conduct.
Interactive
Causes in a structural model
But-for, held fixed, and joint
Method
Testing a harm for dependence
Four steps. The model is written before the test is run, and the verdict belongs to the model.
1
Name the harm and the good
Pick one variable for the harm and one for the good it accompanies. They must be events in the model, not descriptions of the agent.
2
Write the equations from the case
Say what each variable depends on, as the case is described. Whether the good needs the harm is decided here, by what you write.
3
Flip the harm and re-evaluate
Set the harm to its opposite, leave the rest, and read whether the good changes. That is the but-for test.
4
Name the clause tested and the ones not
A harm that is a means fails the clause that the harm not be the means. Whether the act is permissible and the good proportionate are separate questions.
What the flip shows
Score 0 / 0Choose the answer you can justify.
The same five lives, two equations
Both models have the harm true and the five surviving. Flip the harm.
switch: F = E flip H, F stays 1 side effect
loop: F = E and H flip H, F goes to 0 means
same lever, same man, same five
the difference is one equation
In the switch the five are saved by the turning of the trolley, and the man is hit afterwards. In the loop they are saved by the man’s being hit. The lab reports the first as not a cause and the second as a but-for cause, which is the distinction the doctrine draws and nothing more.
Common mistakes
1
Thinking double effect is about intentions the lab can readThe equations contain no mental state. Two agents with opposite intentions in the same loop get identical tiles, because the tiles read what depends on what. Dependence is the observable thing an intention to use the harm is supposed to go with, since a person who sees that the good needs the harm cannot pursue the good without relying on the harm. Whether that is all the doctrine means is part of what is disputed.
2
Treating a side effect as a harm that does not matterIn the switch the man is hit and dies, and the lab’s “not a cause of the good” says nothing about how much his death weighs. The proportionality clause, which the lab does not test, is where that is argued.
3
Reading the verdict off the case instead of the modelThe case does not come with equations. Rewrite F = E as F = E and H and the verdict moves from side effect to means with no change in the world. The classification is only as good as the description it was computed from.
Check your understanding
✓
Finish when you can write a case as equations and say which clause the flip tests.
Given a case with a good and a harm, you should be able to write the structural equations, intervene on the harm, report whether the good depends on it, and name the clauses of the doctrine the test leaves unchecked.
Note: The Boolean models here have no probabilities, so a harm that makes a good merely more likely is not covered. Extending the test to risk is a real open problem for the doctrine, and no lab in this course settles it.