From: Robin Hanson (rhanson@gmu.edu)
Date: Fri Jun 20 2003 - 15:40:30 MDT
At 04:12 PM 6/20/2003 -0400, Eliezer S. Yudkowsky wrote:
>>We were talking about people using self-deception instead of
>>dishonesty. For example, instead of lying and saying you are a great
>>worker, you make yourself believe it even when your evidence goes
>>against this belief. "Bias" mean processes that produce such systematic
>>deviations.
>
>Why should systematic deviations in beliefs be any more or less difficult
>to detect than systematic deviations in statements? Both are information,
>both have probability distributions, both are produced by causal
>mechanisms. What makes one information-theoretically different from the other?
There are four levels that we might examine and compare:
1. Agent statements and actions
2. Agent beliefs
3. Agent internal evidential and other context of belief updating
4. Agent external evidential and other context of belief updating.
If all you have is levels 1 & 4 to compare, you'll have a hard time
identifying a "bias" in a particular case. (Researchers can already
identify such biases on average though, at least in many cases.) The hope
would be that being able to also observe levels 2 & 3 would make it easier
to identify bias. Deviations between levels 1 & 2 are called "lying". And
ways in which level 2 seems to be influenced by non-evidential elements of
level 3 are usually "motivational belief bias".
It was my guess, if you recall, that it would probably be feasible to be
transparent enough to allow outsiders to see lying, but harder to allow
outsiders to see motivational bias. If so, self-deception would become
stronger.
Robin Hanson rhanson@gmu.edu http://hanson.gmu.edu
Assistant Professor of Economics, George Mason University
MSN 1D3, Carow Hall, Fairfax VA 22030-4444
703-993-2326 FAX: 703-993-2323
This archive was generated by hypermail 2.1.5 : Fri Jun 20 2003 - 15:50:13 MDT