Moral uncertainty · working tool

How much worse is the worst future than the best is good?

Spread your credence across moral views, choose how to normalise them, and read off the WORST:BEST ratio R. Cutting the chance of WORST by one percentage point is worth raising the chance of BEST by R points. Extinction is the zero point on every view.

R = [V(extinction) − V(WORST)] ÷ [V(BEST) − V(extinction)]
Start from:
1

How much credence do you give each view?

2

How should CU, NLU and BAUB be normalised?

How likely is it that each view agrees with CU about BEST, rather than about WORST?
Agrees about WORSTAgrees about BEST
Agrees about WORSTAgrees about BEST
Credence in each normalisation method
MethodCredenceIts R
3

Other inputs

4

Splitting or bargaining instead

These don’t give a ratio. Each view controls a share of resources equal to your credence in it, and spends it as its own ratio says. In this model, Nash bargaining with credences as weights gives the same allocation.

Avoiding WORSTPromoting BEST