Negative knowledge values imply reflective or dynamic inconsistency in one’s decision theory. So in each of these cases where you benefit from well-placed ignorance, something in the system has a major value inconsistency. It may very well be optimal for one person within the system to be strategically ignorant, but the system has some crucial flaw nonetheless.
For example, any time you have people deliberately becoming ignorant of an injured person needing assistance, the system is probably excessively punishing (or under-rewarding) those who happen to be nearby. Counterintuitively, then, it may be better to have blanket immunity for those nearby who fail to act so as to increase the chance people will be around at all and who have something to offer.
A similar analysis applies to the trolley problem, much to my consternation: if people regard it as moral to “shuffle” risk around (such as by diverting a trolley from a dangerous to a safe area), then people have to overspend on risk avoidance, thereby reducing the chance that anyone will be around at all to help in such situations, and reducing the total societal resources available to manage risk in the first place.
The negative value isn’t on actual knowledge, it’s on perceived knowledge. Genuine ignorance is advantageous only in the special case where knowledge is difficult to hide.
The same argument applies just the same to perceived knowledge as evidence of a reflective inconsistency in the system. To the extent that someone benefits from spending resources misrepresenting their state of knowledge the system is reflectively inconsistent.
I have about 85% confidence that what you’re saying is correct, but I can’t quite grasp it enough to verify it independently. Maybe I need to reread the TDT paper.
Negative knowledge values imply reflective or dynamic inconsistency in one’s decision theory. So in each of these cases where you benefit from well-placed ignorance, something in the system has a major value inconsistency. It may very well be optimal for one person within the system to be strategically ignorant, but the system has some crucial flaw nonetheless.
For example, any time you have people deliberately becoming ignorant of an injured person needing assistance, the system is probably excessively punishing (or under-rewarding) those who happen to be nearby. Counterintuitively, then, it may be better to have blanket immunity for those nearby who fail to act so as to increase the chance people will be around at all and who have something to offer.
A similar analysis applies to the trolley problem, much to my consternation: if people regard it as moral to “shuffle” risk around (such as by diverting a trolley from a dangerous to a safe area), then people have to overspend on risk avoidance, thereby reducing the chance that anyone will be around at all to help in such situations, and reducing the total societal resources available to manage risk in the first place.
The negative value isn’t on actual knowledge, it’s on perceived knowledge. Genuine ignorance is advantageous only in the special case where knowledge is difficult to hide.
The same argument applies just the same to perceived knowledge as evidence of a reflective inconsistency in the system. To the extent that someone benefits from spending resources misrepresenting their state of knowledge the system is reflectively inconsistent.
I have about 85% confidence that what you’re saying is correct, but I can’t quite grasp it enough to verify it independently. Maybe I need to reread the TDT paper.
Yes, that’s where I got my insight into the reason why reflectively consistent decision theories don’t have negative knowledge values.