Discussion about this post

User's avatar
aubreykohn's avatar

Apparently, "probability estimate" is being used as a more tractable, scalar, proxy for belief.

Expand full comment
Overcoming Bias Commenter's avatar

If Bob and Charlie share the same causal model of some aspect of reality except that the model contains parameters (integers or real number) and Bob and Charlie differ in their estimates of those parameters then yes, I am willing to believe that the average of their estimates is more accurate than either of their estimates.

Since there exists a mathematical theory of causal models (in Bayesian networks and structural equation models) it is possible that you are refering to mathematics which describes a way to average causal models having what one might call "qualitative" differences (different number of equations in the two models, different number of terms in two equations, different number or identity of factors in two terms) rather than merely "quantitative" differences (different integral or real coefficients in the equations). If so, I would love to learn more about this crunchy mathematics.

But if the math you refer to assumes that Bob and Charlie have the same "qualitative" causal model and differ only in their beliefs about the correct values of the parameters of that model, do you really believe that the math is relevant to human beliefs about most things, particularly about aspect of reality as complex as politics?

Expand full comment
78 more comments...

No posts