Re: Operation Politically Homeless

Daniel C. Burton (dburton@ocf.berkeley.edu)
7 Dec 1997 22:11:56 GMT 

Okay, I was thinking about how I could write my computer program to
analyze political persuasions with artificial neural networks, and I
realized how you could make a really fair and accurate political quiz.  I
was thinking about some of the issues with three choices where you can't
reduce it to a yes/no answer and how you could pose a question and then I
remembered those Jungian personality type tests.  In that spirit the
questions would be something like this:

For each question circle the letter you most agree with:

1. a. Government should protect the environment by limiting pollution.
   b. Anti-pollution laws hurt the economy.  Repeal them.
   c. Hold polluters liable for damages to private property.

For your standard 2-choice issues, you would just have two choices.  Each
side would have to appeal in the language of that side.

For example:

a. The government should mandate a living minimum wage.
b. Wage and price controls hurt the economy.  Repeal them.

Of course, if we just wanted to make something objective but also good to
sell libertarianism, we wouldn't want to toss out arguments by the other
side without refuting them.  What we would want to do is prune down the
argumentation and get the questions down to simple statements of position.
We would also have to take care to choose the most important issues.

It looks like it would be very hard work to do this and still keep things
short enough for peoples' attention span.  Since you'd have to be able to
out-do David Nolan at this, this is an argument for just using the WSPQ.

....An artificial neural network on the other hand, I'm guessing could be
accurate with even hopelessly biased questions if the sample size it
learned from was big enough.

Daniel C. Burton <dburton@ocf.berkeley.edu> wrote:

: Wow, all this has inspired me.  I think I'll try to write a computer
: program using artificial neural networks to help people find their
: political home with an arbitary number of arbitary-partite choices on
: issue!  Anyone want to help me?