Please enable JavaScript.
Coggle requires JavaScript to display documents.
181120 Apollo Analysis of forecasting methodology: Film Tetlock (if wrong,…
181120 Apollo Analysis of forecasting methodology: Film Tetlock
if wrong
provide ideological defences
reasons why
counter factual explanations
wrong timing
we made the right mistake
so
we need standards for judging foreasts
need to record good judgement consistenty
experiment
questions
tight enough to avoid clavoyance
found
traditional views don't make much difference
content was not a good predictor
what mattered how forecasters fought
foxes over hedgehogs
foxes know many things
persue many ideas without one embracing visions
reluctance to predice
hedgehog knows one big thing
relates everything to one lens / a deductive frame of mind to first principles to a framework
tend to be more confident in their abilities
but assign higher probabilities to low frequency events
but the price is high positives
2,3 times more likely to assign probabilities of 0s or 1s
people respond to 0s or 1s but generally not to small movements within the range
usually do better when aggregated
but not a horse race
but in an interdependant system
foxes as intellectual scavencers
so should be considered when forming teams
some genetic and some upbringing in determine hedgehog or fox
about 20 % foxes, 20 % hedgehogs rest lie somewhere in between
self contrarian style may led to realistic possibilities to positive futures
but everyone gets some things right
the very best foxes are only slightly better than simple algorythms
but only marginally better than others
aggregates of experts do better than individuals
take aways
aggregation improves forecasting
hedgehogs over foxes
inverse relationship between charisma and accuracy
can people be trained?
potentially some generalised ability
Marcus Butler
getting better because you are closer to the event is not valuable