[EM] Use of "bit"

[Forest Simmons]

I propose that we use "bit" in the standard computer science usage, namely
the number of yes/no or on/off or marked/blank options available for
codifying information.

So an n bit style ballot allows (at maximum) 2^n different options
(constraints lacking) per candidate.

In particular, a three bit ballot allows eight options, enough to provide
the seven levels of distinction that are about right according to
cognitive psychologists.

[Josh brought this up recently, and we discussed this in a similar context
last year under the heading "seven +/- two."]

So, for example, three bits are adequate for both seven slot approval and
cardinal ratings with resolution seven.

A name on a ballot followed by three ovals to mark, e.g.

Jane Candydate  (4) (2) (1)

allows eight options, namely the sum of the numbers inside the marked
ovals, a value ranging from a minimum of zero (no mark) to seven (all
marked).

Even Demorep's Joe Q Public can reliably add numbers in this range.


To distinguish three levels it takes two bits of information, enough to
also distinguish four levels, so the recently discussed three level
"Majority Choice Approval" variants are intentionally wasting one level to
err on the side of simplicity.

Thus

Jane Q Candydate  (X) ( )

and

Jane Q Candydate  ( ) (X)

are both interpreted as approval for JQC, but not as favorite status,
which would look like this:

Jane Q Candydate  (X) (X)


If standard voting machinery and standard one bit style ballots are
required, any two bit method can be implemented by doubling all of the
names on the ballot:

Jane Q Candydate  ( )
Jane Q Candydate  ( )
Jill R Running    ( )
Jill R Running    ( )
etc.

This would have a good side effect; the redundancy would make it more
likely that at least one of your marks got next to the intended name. Even
the butterfly ballot would be harmless if the names across from each other
were identical.

Tripling for three bits would be possible, but bordering on the
impractical and unwieldy.

Forest

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em