[EM] danger of coercion (Re: First U.S. Scientific Election Audit...)

[Abd ul-Rahman Lomax]

At 06:32 AM 5/17/2007, Juho wrote:
> >> What would be the most likely scenario where the coerced person could
> >> not avoid being coerced? I'm assuming that a typical coercer would
> >> not be a member in the team that counts the votes and he/she would
> >> not have open access to the ballots.
> >
> > That's not reliable.
>
>I asked for the most likely scenario. I'll take this to mean that
>maybe the most likely scenario is one where the coercer sits in the
>vote counting team.

The victim does not necessarily know where the coercer, or someone 
allied with the coercer, sits. I don't think, in addition, that we 
can make any general statement about how likely it is that the 
coercer has inside access. We can, indeed, conclude that access to 
what is visibly expressed on ballots will broader the actual access 
of one who would coerce votes, but it also will make coercion schemes 
more visible and more likely to be discovered. I don't think we can 
predict whether ballot imaging will, overall, make coercion more or 
less likely. My intuition is less, but I can easily understand that 
someone would think that it would be more.

However, the level of cooperation with coercion is very unlikely to 
be large; if it looks like it is headed that way, procedures can be 
revised. *Massive* coercion is actually easier to detect and defeat, 
what would be very hard would be the isolated coercion of one 
individual over another, such as a spouse coercing a spouse. I think 
this, however, would remain extremely rare. And vote coercion should 
be treated as a serious crime. It is a dangerous business for the 
coercer, actually, much more than for the coerced. If my vote is 
coerced, the cost for me to comply is small. It is really only when 
large numbers of votes are coerced, in some pattern, that a different 
level of cost emerges. One vote is only one vote, it is vanishingly 
rare that it affects an election outcome. As I mentioned, if evidence 
emerges, as it must with public imaging, that there is more than 
minimal coercion, steps can be taken to interdict it. Those steps 
have a cost. One of the obvious steps is to shut down the imaging 
program. I consider that a high cost, frankly. There would have to be 
more than a rare instance of coercion to make the disease less costly 
than that particular remedy.

>An alternative method is to require potential additional candidates
>to collect a list of e.g. 1000 supporters before the election and
>thereby become "regular candidates".

My own suggestion has been merely to require registration. The 
proposal Juho makes misses the point of write-in candidacies. They 
are for candidates who were unable to get on the ballot. Ballots are 
printed in large numbers, with fairly onerous security, and they must 
be widely distributed. They cannot be printed the day before the 
election, it would practically guarantee that some polling places 
would not get their ballots. Presses can break down, printing can be 
delayed. You don't want to push it. There are absentee voters as 
well, who need ballots in advance of election day.

Registration would result in the candidate receiving a registration 
number, which could be indicated on the ballot, where needed, using 
standard marks, avoiding the written name, which would be far more 
reliable for identification.

But my real point is that we don't need cumbersome restrictions to 
solve a problem that is practically nonexistent. We are trying to 
avoid coerced voting where the coercer requires the voter to make the 
ballot identifiable. Quite simply, I expect this to be vanishingly 
rare. When you get millions of people voting, "rare" may be almost 
guaranteed to happen sometimes. But that does not mean that we stand 
everything on its head to prevent a rare occurrence. Rather, we 
consider the cost of that occurrence and balance it with the cost of 
attempting to totally prevent it.

My advice to someone who is a victim of attempted vote coercion 
requiring validation? If you fear that your vote will actually be 
observed, that if you do not mark your ballot as required so that it 
satisfies the coercer, you will be subject to serious harm, vote as 
required. And if you can find any authority you trust, report that 
you are doing so. Your vote is visible, by the conditions of this 
problem. You can prove that you voted in this way. And you could, for 
example, wear a wire when you go back to your dear tyrant husband and 
tell him that "I voted as you wished, dear, but I'm worried that the 
mark I made will rub off." If he answers "You better hope it 
doesn't," then, well, you can be relieved of his presence if that is 
what you want.... otherwise ....

Of course, if you have enough time, you could have a similar 
conversation before the actual election....

But if you think all this too risky, and it's your call, just vote as 
directed. The cost to society is only one vote, and your life and 
security are far more important than that.

>I propose simple ballots and separate ballots for each race in
>addition to what you said.

Not a snowball's chance in h e double toothpicks, as my wife's father 
used to say.

The public expense is relatively large. What is the evidence that the 
cost of not doing this is greater?


> >> Of course cameras and video equipment should be banned in the voting
> >> location. It is not possible to guarantee 100% that such recording
> >> will not take place but one should try.
> >
> > I disagree. And nobody searched me for my cell phone when I last
> > voted.
>
>No need to remove cell phones, just to make it clear to all that
>taking pictures is forbidden (maybe even punishable if extreme
>measures are needed).

Oh, dear. But vote coercion is already a serious crime. Why would 
such a rule make any difference at all?


> >> Combining multiple elections in one ballot is a risk.
> >
> > Sure. But it is absolutely the norm. In fact, I've never voted in a
> > public election where there were not many races on the single ballot.
>
>Bad design. I have never voted in an election with several races on
>one sheet. (Usually there has been only one race per election, but
>when there have been more the ballots have been separate.)

Really? Where do you live? What elections have you participated in?


> >> The risk of allowing access to the ballots to everyone is much
> >> riskier than having multiple vote counters (maybe not local people)
> >> each counting a small portion of the votes.
> >
> > You say so. What is the evidence?
>
>No evidence, just the understanding that allowing the ballots to be
>inspected by whoever has interest, with sufficient time to do careful
>analysis and with whatever techniques may reveal something of the
>identity of the voters of the ballots.

If the voters do not choose to make that information explicitly 
visible, it would be extremely impossible simply from looking at the 
images. If you can gain access to the ballots themselves, it is more possible.


> > There is a lost performative in the last comment from Juho. "In
> > having multiple vote counters"....
> >
> > *Who* is going to "have" multiple vote counters?
>
>The society. Sorry for my non-native English.

"The society" has the same problem. To say "the society" avoids 
specifying who actually counts the ballots. It avoids facing the 
issue of how counters are chosen, and how that process can be 
manipulated. If you *really* want the society to count the ballots, 
you might warm to my idea! That is exactly what I'm proposing:

> > The fact is that this is what I'm suggesting: multiple vote counters!
> >
> > Ballot imaging would put the public in the position of being an
> > election observer, to a degree.
>
>The step of initial vote counting (and possibly imaging) may be even
>more critical point to safeguard.

What I'm suggesting is this: votes are typically deposited in a 
ballot box. In the small town where I recently lived, the ballot box 
was a machine that is probably over a hundred years old. It has a 
slot on the top and a crank that you turn when you insert a ballot in 
the top. The machine is designed such that you can insert a ballot 
into it, but you cannot retrieve a ballot from it, without removing 
the locks on a door. It is not always that elaborate, but the point 
is that ballots are deposited into a secure box that is guarded. In 
fact, these things, the boxes, get lost, or are deliberately 
discarded in some places, but that is another story.... It would be 
pretty hard to lose the box in Cummington, it was pretty bulky....

Anyway, at a certain point the box is opened and the ballots are 
removed. I'm suggesting that this would take place with multiple 
observers, and that the ballots would be immediately numbered, 
imaged, and sequestered. When they are being imaged, they would be 
available, under continuous observation, for imaging by observers. 
Certainly an observer from each party, but also media observers. This 
all happens together, in a place where everyone sees what everyone 
else is doing. The handling of the ballots would be closely 
controlled, but the best control is redundant observation of every step.

The counting is then done with images, not with the original ballots. 
The counting is *easier* than now, because the security required is 
much less. The images are expendable and replaceable. The counting is 
done with a copy of the images, it is not as if there is only one 
image and they handle it. That's the official count. It can, with 
images, be done more rapidly than with actual ballots. But parallel 
to this, media and other interested individuals would also be 
counting the ballots, using images made available on the internet. 
I'd personally find it interesting to count a few.

That's what the serialization is about. When I count a ballot, I 
would enter the information on a form. If we assume that the serial 
number is easily machine readable, that number could be automatically 
filled in on the form; and so all I would need to do is check off 
"votes" on a form that resembles the ballots, converting the image to 
vote data. At the same time, I'd assume that automated recognition 
would be doing the same thing. Anyone could run the automated 
recognition software, so I'm not sure how much actual hand counting 
would take place. Public databases would be built up with all this 
information. There will be count conflicts, and these will be 
flagged, and then people will pay special attention to those ballots. 
Truly ambiguous ballots will be identified.

So a final report, which would represent a general consensus, would 
segregate ambiguous or controversial ballots from those where there 
is no controversy. None of this public process affects the official 
outcome, which is announced from the official counting process. 
Rather, the public process is a watchdog, it barks if there is 
something wrong with the official count (which likewise reports exact 
ballot information as entered, including the serial numbers). The 
public count does not interfere with the official count.

If however, the public count shows discrepancies with the official 
count, there are then immediate and clear grounds for a recount 
request, and clear evidence to present, not speculation. And if the 
recount leaves the discrepancies unresolved, then there would be 
court action, and it might be necessary to examine the sequestered 
ballots. I'd expect this to be extremely rare.

> > By the way, ever try to get correlated election data? Did people
> > who voted for Bush vote against the school bonds? That kind of data
> > has plenty of legitimate use. And you can't get it now, unless you
> > are willing to spend prohibitive amounts of money to get it....
>
>Sounds a bit dangerous from privacy point of view. The next step
>could be to include the voter's profession, age etc. Better be
>careful with these.

Juho, nothing was suggested about collecting private information. 
What was suggested was that the information on ballots, which has 
become *public* information, in fact, be made available. It's already 
public information, you should understand that. You can get access to 
it if you want. But it is inordinately expensive. I'm not sure how 
much the media spent recounting in Florida after the 2000 election, 
but it was a fortune. It was a consortium that did it.

If ballot imaging had been in place, and the ballots had been simple 
paper ballots, the whole thing would have been over in a few days. 
For the public to count the entire election with massive redundancy 
should not take more than a day....

One of the big errors made in Florida was to require that a 
*decision* be made on each ballot. Because of the nature of those 
voting machines, there were lots of ambiguous ballots, with the 
famous hanging chad. Instead of deciding on each ballot, the court, 
in my opinion, should have ordered and instituted a classification 
system. A set of observers would follow established standards for 
identifying ballot characteristics. This was done to some degree, but 
instead of requiring a decision on the vote result right then, I 
would have had, instead, the *votes* of the observers recorded as to 
the characteristics. The court would have then sorted it out by 
determining the exact standards to apply: for example, a two or three 
corner attached chad might be considered unpunched, and a one-corner 
as having been punched. Specific overvotes would have been reported 
instead of merely discounting the ballots and recording them as 
spoiled. Etc., etc.

It's really what any scientific investigation would do....