3 ms·

You can slightly change the context of the problem, and it entirely changes the character of responses: Imagine you're an accountant working for the government

You can slightly change the context of the problem, and it entirely changes the character of responses: Imagine you're an accountant working for the government. You choose how medical research funds are spent. One day, you notice that there are two equally expensive programs: Program A, which will save one life next year, and Program B, which could save five lives next year. Program A is being funded, but Program B isn't. Program B could be funded if you took all of the funding from program A and diverted it. Do you divert the funds?

sokoloff 11y ago
> two equally expensive programs: Program A, which will save one life next year, and Program B, which could save five lives next year. For me, it depends on the chances of "could save five lives" and the cost of the program. If the chance of "could save 5" is under 20%, no, you don't divert the funds. If the total expense of the program is over some threshold per life saved, and there is another opportunity, C, elsewhere to expend those funds with a better outcome (more lives saved/extended/improved), you defund both A & B in favor of that other program.
- barrkel 11y ago
  I would expect that probability has already been factored into the numbers at this level of decision. If you're e.g. making a change to road signage, you wouldn't start out with the total number of KSIs, you'd be looking at the potential reduction in KSIs. Similarly, if you're looking at a medical intervention, you'd be measuring a probability-adjusted delta in QALYs (in the UK).
  - sokoloff 11y ago
    Yes, I agree that's how it should be done. The question specifically used certain language for one branch and uncertain language for the other, so I replied to the question as asked rather than strawmanning an answer to a question that wasn't asked.
    - chc 11y ago
      I think that's because one was a statement of present reality (the program is currently doing this), whereas the other was a hypothetical (if could do this if we funded it).
      - sokoloff 11y ago
        Thanks. I now see that as a possible reading. I read it only as a statement of uncertainty. s/could/would/ would have eliminated the ambiguous reading.
- nickodell 11y ago
  Ack, sorry. The two programs are equally likely to save lives, but one is currently funded, and the other isn't. That's the reason for the would/could distinction.
lm______ 11y ago
This scenario is slightly different though. It doesn't explicitly preclude the possibility that if funds are diverted to program B, that the one person being saved by program A might also be saved under program B, plus an additional four others.
jasonkostempski 11y ago
But there are 2 parts to the thought experiment. This seems just like the "flip the switch" scenario, which most people say they would do.
Retra 11y ago
There is a clear inefficiency in diverting the funds. It takes effort and money to change an already-active process.
r00fus 11y ago
What's the cost of switching? What are the respective probabilities of success? What about the possibility that a switch might lead to both projects failing? Distilling real-life into utilitarianism scenarios is fraught with reductionism - there are many factors that drive people's decisions and those all add up.

sokoloff 11y ago

> two equally expensive programs: Program A, which will save one life next year, and Program B, which could save five lives next year. For me, it depends on the chances of "could save five lives" and the cost of the program. If the chance of "could save 5" is under 20%, no, you don't divert the funds. If the total expense of the program is over some threshold per life saved, and there is another opportunity, C, elsewhere to expend those funds with a better outcome (more lives saved/extended/improved), you defund both A & B in favor of that other program.

barrkel 11y ago
I would expect that probability has already been factored into the numbers at this level of decision. If you're e.g. making a change to road signage, you wouldn't start out with the total number of KSIs, you'd be looking at the potential reduction in KSIs. Similarly, if you're looking at a medical intervention, you'd be measuring a probability-adjusted delta in QALYs (in the UK).
- sokoloff 11y ago
  Yes, I agree that's how it should be done. The question specifically used certain language for one branch and uncertain language for the other, so I replied to the question as asked rather than strawmanning an answer to a question that wasn't asked.
  - chc 11y ago
    I think that's because one was a statement of present reality (the program is currently doing this), whereas the other was a hypothetical (if could do this if we funded it).
    - sokoloff 11y ago
      Thanks. I now see that as a possible reading. I read it only as a statement of uncertainty. s/could/would/ would have eliminated the ambiguous reading.
nickodell 11y ago
Ack, sorry. The two programs are equally likely to save lives, but one is currently funded, and the other isn't. That's the reason for the would/could distinction.

lm______ 11y ago

This scenario is slightly different though. It doesn't explicitly preclude the possibility that if funds are diverted to program B, that the one person being saved by program A might also be saved under program B, plus an additional four others.

jasonkostempski 11y ago

But there are 2 parts to the thought experiment. This seems just like the "flip the switch" scenario, which most people say they would do.

Retra 11y ago

There is a clear inefficiency in diverting the funds. It takes effort and money to change an already-active process.

r00fus 11y ago

What's the cost of switching? What are the respective probabilities of success? What about the possibility that a switch might lead to both projects failing? Distilling real-life into utilitarianism scenarios is fraught with reductionism - there are many factors that drive people's decisions and those all add up.