Most value is in knowing what's going to happen
We ask experts for this "what will this policy do?" "will the space rocket fly?"
For political predictions (which most predictions that aren't already predicted by the stock market) experts are worse than random
Well-calibrated forecasters are good at this
This seems to imply a better system would be to get decision-makers to ask groups of forecasters what will happen and for them to ask the experts if they need domain expertise
This has two kinds of problems
It is elitist. Elites always like to contrive of reasons that they alone should get to make the decisions
It is opaque. If the forecasters seem to be biased, what is the response?
For these reasons it might have led to atrocities in the past. Elite groups have supported communism, fascism, development of nuclear weapons, etc etc. Elites are frequently poor at making decisions for the whole
Alongside this process therefore I would like another, which is more transparent
People create probabilistic models of their understanding of the situation eg
EstimakerPeople input their own probabilities into other people's models
People choose between different models
For the moment, let's say voting, but equally it could be
Trying to build models that are supersets of all other models (too intensive to fill in)
Dictator (me) makes models that get to the cruxes of other models (again, elitist & opaque)
Something more elegant
Where they disagree on either probabilities or models, there can be narrow, focused debate (see
Better Debates)
Create models again and see if there is more concensus
Test models for accuracy
If there is both accuracy and consensus and transparency then this seems an alternative that is better than the current systems