[PREV - SPINOZA_CYCLE] [TOP]
MEDIUM_SPEED
January 01, 2014
Daniel Kahneman, "Thinking, Fast and Slow" (2011), January 05, 2014
Ch 21 "Intuitions vs Formulas"
p.227 or so (hardback)
Kahneman gets really interesting at this stage. FAST_SLOW_AND_SLOWER
He's beginning to make prescriptions on how to
make decisions given the limits of the two
"systems".
He presents an insight from the world of medical diagnosis:
mechanically running through a checklist often works better
than going by the impressions of the medical staff, no
matter how well trained their expert intuition.
See:
"A Checklist Manifesto"
Atul Gawande
There's a long and personal (for him) anecdote at
the end of this chapter, of working out methods of
predicting the success of soldiers in the Israeli
army in 1955.
He settles on an approach of looking for detailed,
easily measurable parameters (typically around 6) Note: some aspects
that can be combined in a simple formula to get a of this approach are
prediction. proven, others
perhaps less so.
Two features are of interest:
(1) the weighting of the parameters can be equal:
parameter weights are over-rated, and
don't help as much as is often assumed.
(2) he does not exclude the examiner's inuition,
but treats it as another input to be weighted.
He presents an accidental discovery (how well
confirmed, I'm not sure at present), that an
interviewers intuition works better if they're
first asked to go through a standard set of SUPERFORECASTING
detailed questions.
So that's another area of interest:
what can you do to improve your intuition?
SUPERFORECASTING
My attempt at firming up the
process Kahneman discusses:
o first use your intuition to identify
parameters that are likely to apply.
o make no effort to determine
relative strengths: if it's in it's in.
o include one last parameter,
the educated guess of an expert observer.
o determine a predictive factor from this,
and over time make observations to check
how well the predictions have worked.
Note that this relies on certain observable
inputs, and a well-defined output.
One thing I don't quite follow from Kahneman's
account: they were interested in predicting
a candidate's future performance as a soldier: The difficulty: many of
how was that measured to see how well their the things they're
technique worked? testing might be affinity
tests in disguise, and if
(Could it be that Kahneman has the "success" parameter
overapplied techniques that work well is similarly skewed it
with medical problems, but don't map could be this grand
quite that well to social ones?) result is effectively "we
like people like us".
This approach allows a role for
intuition, but it (a) tries to
guide it and (b) it's a strictly
limited role:
Kahneman suggests that six parameters
is typically a reasonable number to
work with, and treating intuition as
just one more parameter puts it at
around 15% of the decision.
Except of course, that the
selection of the other parameters
is also a matter of intuition.
--------
[NEXT - OPTIMAL_MIST]