Re: [EM] Open Voting Consortium for e-voting?

[Ernest Prabhakar]

On Apr 16, 2004, at 2:12 PM, Ken Johnson wrote:


From: Ernest Prabhakar <[EMAIL PROTECTED]>
Date: Fri, 16 Apr 2004 11:12:01 -0700
...
...  Is there any way other groups could  get access to summaries 
(e..g, the pairwise matrix) or scrubbed subset  of ballots, for 
research purposes.

Certainly precinct totals should be made available.  We haven't done  
the threat analysis to determine whether or not individual ballots  
should be made available.

In my view, public disclosure of precinct totals should be construed 
as a potential violation of voter secrecy. If a high percentage of 
people in my precinct vote a particular way, one could draw a highly 
probable inference about how I voted based on the precinct totals. 
Politicians may show favoritism toward or bias against certain 
precincts based on their voting patterns, or may use the information 
for targeted political marketing campaigns.

Furthermore, public disclosure of ballots, even devoid of any voter or 
precinct identity information, could also be construed as a secrecy 
violation because the ballots could be used to determine correlations 
between different ballot issues. For example, if the election includes 
a ballot initiative for Electoral Reform, one could use the ballots to 
determine the political constituencies of Electoral Reform advocates. 
This kind of information might make some politicians less likely to 
support Electoral Reform.

Of course, the ballot information could be obtained by conventional 
polling methods (e.g., exit polling), but in this case the voters have 
the option of withholding or misrepresenting their views. From my 
perspective, the only kind of "research purpose" for which ballots can 
be legitimately used is to determine the election outcome. Ballots and 
precinct-level totals should only be available to election officials 
and auditors who are obligated to maintain such information in 
confidence.

Ken Johnson
From my perspective, what would be most useful for research is 
-aggregate-  data across the entire electorate broken down by contest.  
 For plurality, that's pretty much what we get already, so its not all 
that interesting.  For ranked-order ballots, its a lot more 
information, but still it should be pretty straightforward to compile 
and release.

If broken down by contest, that should avoid the correlation concerns. 
Would that work for you?

-- Ernie P.



----
Election-methods mailing list - see http://electorama.com/em for list 
info
----
Election-methods mailing list - see http://electorama.com/em for list info