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).
Re:The Cold Logic of Drunk People (2014)
by sokoloff on Monday October 26, 2015 @08:16AM
(#10450846)
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.
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).
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.