From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Tue Jun 17 2003 - 15:46:14 MDT
Robin Hanson wrote:
>
> Yes, humans are naturally moderately rational, and naturally use and
> improve that rationality as long as doing so does not threaten our
> cherished self-deceptions. The extra cost I had in mind was having to
> deal with an "enemy" (our subconscious) that knows us very well and has
> had long evolution-embodied experience in fooling us into thinking we
> have made progress when we have not. My estimate that this cost is very
> high is in part based on the very small fraction of people that seem to
> incur this cost and successfully overcome their biases.
This cost is not a loss as a result of rationality, it's an expenditure
that needs to be made to get to rationality... um, review: what is it, if
anything, that we still disagree about? I would say that, even
instrumentally, the benefits of rationality are higher than the losses,
and that the costs involved do not alter this. I am under the impression
you still disagree with this, or that you assign it a weaker degree of
belief? I confess I wasn't even counting the time-energy budget as a
loss, just taking it for granted as something a person's gotta do
eventually. However, even taking into account the substantial cost in
time and mental energy, I still say it's the all-time best investment I
ever made and I still think it's worth a shot for others too.
> As an aside, I actually have high hopes that we can improve people's
> incentives to be rational in their contributions to collective consensus
> via wider use of betting markets. People are more rational when they
> bet, for obvious reasons.
Depends on the bias. If I recall the research correctly, most biases
studied don't go away because of minor financial incentives. Whether an
entire market with major incentives can, over time, build up a discipline
and a body of knowledge to beat bias, is a separate issue. But, yes, I
agree that rationality may be substantially improved.
-- Eliezer S. Yudkowsky http://singinst.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Tue Jun 17 2003 - 15:56:03 MDT