“Deserve” is harmful because we would often rather destroy utility than allow an undeserved outcome distribution. For instance, most people would probably rather punish a criminal than reform him. I nominate “justice” as the more basic bad concept. It’s a good concept for sloppy thinkers who are incapable of keeping in mind all the harm done later by injustices now, a shortcut that lets them choose actions that probably increase utility in the long run. But it is a bad concept for people who can think more rigorously.
A lot of these “bad concepts” will probably be things that are useful given limited rationality.
“Are the gods not just?”
“Oh no, child. What would become us us if they were?”
I’d say “justice” is a heuristics; better than nothing, but not the best possible option.
For instance, most people would probably rather punish a criminal than reform him.
This could be connected with their beliefs about probability of successfully reforming the criminal. I guess the probability strongly depends on the type of crime and type of treatment, and even is not the same for all classes of criminals (e.g. sociopaths vs. people in relative rare situation that overwhelmed them). They may fear that with a good lawyer, “reform, don’t punish” is simply a “get out of jail free” card.
To improve this situation, it would help to make the statistics of reform successes widely known. But I would expect that in some situations, they are just not available. This is partially an availability heuristics on my part, and partially my model saying that many good intentions fail in real life.
Also, what about unique crimes? For example, an old person murders their only child, and they do not want to have any other child, ever. Most likely, they will never do the same crime again. How specifically would you reform them? How would you measure the success of reforming them? If we are reasonably sure they never do the same thing again, even without a treatment, then… should we just shrug and let them go?
The important part of the punishment is the precommitment to punish. If a crime already happened, causing e.g. pain to the criminal does not undo the past. But if the crime is yet in the future, precommiting to cause pain to the criminal influences the criminal’s outcome matrix. Will precommitment to reforming have similar effects? (“Don’t shoot him, or… I will explain you why shooting people is wrong, and then you will feel bad about it!”)
I nominate “justice” as the more basic bad concept. It’s a good concept for sloppy thinkers who are incapable of keeping in mind all the harm done later by injustices now,
Actually, I think that’s some of what they are keeping in mind and find motivating.
If they were able to keep it in mind separately, they could include that in their calculations, instead of using justice as a kind of sufficient statistic to summarize it.
You can still use precommitment, but tie it to consequences rather than to Justice. Take Edward Snowden. Say that the socially-optimal outcome is to learn about the most alarming covert government programs, but not about all covert programs. So you want some Edward Snowdens to reveal some operations, but you don’t want that to happen very often. The optimal behavior may be to precommit to injustice, punishing government employees who reveal secrets regardless of whether their actions were justified.
International espionage is probably one of the worst examples to attempt to generalize concepts like justice from. It’s probably better to start with simpler (and more common) examples like theft or murder and then use the concepts developed on the simpler examples to look at the more complicated one.
“Deserve” is harmful because we would often rather destroy utility than allow an undeserved outcome distribution. For instance, most people would probably rather punish a criminal than reform him. I nominate “justice” as the more basic bad concept. It’s a good concept for sloppy thinkers who are incapable of keeping in mind all the harm done later by injustices now, a shortcut that lets them choose actions that probably increase utility in the long run. But it is a bad concept for people who can think more rigorously.
A lot of these “bad concepts” will probably be things that are useful given limited rationality.
“Are the gods not just?”
“Oh no, child. What would become us us if they were?”
― C.S. Lewis, Till We Have Faces
I’d say “justice” is a heuristics; better than nothing, but not the best possible option.
This could be connected with their beliefs about probability of successfully reforming the criminal. I guess the probability strongly depends on the type of crime and type of treatment, and even is not the same for all classes of criminals (e.g. sociopaths vs. people in relative rare situation that overwhelmed them). They may fear that with a good lawyer, “reform, don’t punish” is simply a “get out of jail free” card.
To improve this situation, it would help to make the statistics of reform successes widely known. But I would expect that in some situations, they are just not available. This is partially an availability heuristics on my part, and partially my model saying that many good intentions fail in real life.
Also, what about unique crimes? For example, an old person murders their only child, and they do not want to have any other child, ever. Most likely, they will never do the same crime again. How specifically would you reform them? How would you measure the success of reforming them? If we are reasonably sure they never do the same thing again, even without a treatment, then… should we just shrug and let them go?
The important part of the punishment is the precommitment to punish. If a crime already happened, causing e.g. pain to the criminal does not undo the past. But if the crime is yet in the future, precommiting to cause pain to the criminal influences the criminal’s outcome matrix. Will precommitment to reforming have similar effects? (“Don’t shoot him, or… I will explain you why shooting people is wrong, and then you will feel bad about it!”)
Actually, I think that’s some of what they are keeping in mind and find motivating.
If they were able to keep it in mind separately, they could include that in their calculations, instead of using justice as a kind of sufficient statistic to summarize it.
Would you also two box on Newcomb’s problem?
You can still use precommitment, but tie it to consequences rather than to Justice. Take Edward Snowden. Say that the socially-optimal outcome is to learn about the most alarming covert government programs, but not about all covert programs. So you want some Edward Snowdens to reveal some operations, but you don’t want that to happen very often. The optimal behavior may be to precommit to injustice, punishing government employees who reveal secrets regardless of whether their actions were justified.
International espionage is probably one of the worst examples to attempt to generalize concepts like justice from. It’s probably better to start with simpler (and more common) examples like theft or murder and then use the concepts developed on the simpler examples to look at the more complicated one.