Fog Creek Software
Discussion Board




Voting Machine Usability

I'm interested to hear what people thought about the new voting machines used in some (U.S.) precincts today, from a usability perspective.

Specifically:
- did the voting software follow your expected “voting workflow”?
- were they easier to use than punch cards?
- will they prevent another fiasco like the 2000 election?
- are they secure enough to prevent voter fraud?

Note:  THIS IS NOT A POLITICAL TROLL!  Save your comments about "<insert party affiliation> are too stupid to vote" for elsewhere.

I don't know if the software and touch screens were the same in all states.  My comments are based on the systems used in Georgia.

1.  There needed to be a confirmation for the "Cast Ballot" function.  Normally I hate the "are you sure you want to do this?" dialogs, but in this case I think it's justified.  Pressing this button is final, and there's no way to get another ballot card.

2.  After paging through the ballot entries, you get a chance to review your choices on one page.  Your choices are listed in a scrollable area, with the vertical scroll bar on the right.  Unfortunately, the above mentioned "Cast Ballot" button is positioned just below the scroll down button of the scroll bar.  I can see lots of voters with shaky hands accidentally hitting the "Cast Ballot" button when they meant to continue reviewing their ballot.

3.  Why only one ballot card?  In the past, you could get another punch card if you screwed up.  Is it because of the cost of the cards?  If so, that’s another justification for point #1.

4.  I think the end result of the system should be a printed ballot.  It makes me very nervous that my ballot is stored on what looks like an access card for a satellite TV receiver.  There should be some kind of paper trail.

Joe Paradise
Tuesday, November 05, 2002

My biggest complaint is the lack of privacy.  Ballots are supposed to be secret.  With the electronic machines there are a couple of side panels to limit viewing, but it is not too difficult to look over someone's shoulder.  There was one machine set up to one side and on a lower table for those that needed access from a wheel chair.  While waiting in line it was easy to see which selections were being made.  Although the text was hard to read from a few feet away, it would probably be possible to correlate item selected with the layout pattern of choices.

I agree with your concern about the cast ballot button.  It is too close to the scroll button.  With the old machines you only got one chance, but it was clear what you were doing when you grabbed the big lever.

mackinac
Tuesday, November 05, 2002

Just voted here in Seattle, and while we don't have a software-based solution, the overall usability seemed pretty good to me.

The ballots were scantron-type "fill in the little bubble" forms with clear descriptions of each ballot measure and/or candidate, so it would be very difficult to fill in the wrong bubble by mistake. And if you do think you goofed, you can ask for a new form.

The only thing I saw that seemed potentially confusing is that if you want to specify a write-in candidate, you need to write in his or her name AND fill in the bubble next to the name.

When you're finished, you feed the form into a little reader (looks kind of like the automatic document feeder on a photocopier) and it drops it into a secure box. An LCD on the front of the reader increments its count to show that your ballot was counted.

I don't know whether the reader actually interprets the marks on the spot or just records and saves the ballots for later tallying, but my overall impression was that the system seemed straightforward.

I was a little surprised that I didn't have to show any kind of ID to vote, just tell them what precinct I was in, give them my name, and sign the ledger. Somehow it seems to me like positive ID should be required to help reduce voter fraud, but I guess there's some reason it's not (maybe to avoid putting too many hurdles in the way of people's ability to vote?).

John C.
Tuesday, November 05, 2002

Here (Seattle) they switched to a fill-in-the-bubble paper system a year or two ago. I see no way to significantly improve on it.

a) Paper trail. Obviously, you're voting with paper.
b) Ease of use. Paper is easy to use. The questions could sometimes be in a better format, but they print a summary right there on the page and you fill in a bubble.
c) Random access. It sounds like your touch screen version is like a wizard with verification at the end, very hard to answer unless you've prepared a cheat sheet. (Was that yes or no for Referendum 1234?)
d) Verification. You feed your own ballot into the counter, and watch it approve or reject it. You could still have voted for the wrong person, but gross mistakes are rejected. You can also request a second ballot if you mess up.
Note on verification: some places have this system with the 'reject bad ballots' option turned off. There was lots of suspicion of this in places like Florida, there should be zero bad ballots when the system isn't misconfigured.
e) Distribution. You get your ballot, then fill it in in a little booth. One electronic machine, a lot of little desks with curtains, and if they're full you can fill in the ballot anywhere.
I don't know about accessibility--if you're blind, do they have a second version of the ballot, or have someone (hopefully impartial) assist you?

As for checking ID, I'm not sure what Washington law is, they don't check. In Pennsylvania they cannot check, and the current legislation to require checks is encountering resistance--ID is not a legal requirement to live in the US, and many seniors (who vote) don't have good ID. Of course dead people (who sometimes also vote) don't have good ID either. They do cross-match your ballot with your name, which isn't exactly secret, though I think they rip off the ID part before handing it to you.

mb
Tuesday, November 05, 2002

The main problem I have with paper ballots is that the technology just doesn't seem precise enough for counting something so important.  Obviously nothing can stop people from voting for the wrong person... I'm talking about partially filled in bubbles, etc.  I guess I just have a strong distrust for "scantron" type devices after years of schooling where they often made mistakes.

Mike McNertney
Tuesday, November 05, 2002

I think all the new electronic polling machines are much worse than filling in the bubbles. I don't think the electronic version will match the accuracy of bubble-filling with respect to representing the voters wishes in our life-times. Filling in bubbles is much, much easier to do accurately than using a computer for the vast majority of the voting population.

pb
Tuesday, November 05, 2002

One advantage to the new computerized system in Georgia is that the Secretary of State is publishing real-time election results as the ballots are counted.

http://www.sos.state.ga.us/elections/results/2002_1105/default.htm

Do "fill in the bubble" systems have that capability?

Joe Paradise
Tuesday, November 05, 2002

Actually, Joe, real time results can be big disadvantage - people who plan and run elections tend to be concerned about information revealed influencing peoples' voting patterns.

Rodger Donaldson
Tuesday, November 05, 2002

Sure, the bubble machines can return data just as fast as any other computerized system--it's a scantron machine.
But a voting machine probably should not be examined closely until the polls closed, for obvious reasons--we're about to lose, go round up the dead people in a small town in Texas.

The other big advantage is that it's paper--enough people claim their votes were counted wrong, pull out the papers and recount them all by hand. Enough people claim that the one-pixel sepearation between yes and no wasn't enough, too bad.

Glenn Reynolds (instapundit.com) happens to be discussing these issues a lot today.

Can anyone name any advantages to touch screen type system? The only two I can think of is smaller paper usage, and infiniate availablity of multiple ballots (e.g. English, Chinese, Icons, replacement of issues past the deadline). Some of these 'advantages' are more bugs than features.

mb
Tuesday, November 05, 2002

I don't trust these touch-screen voting machines one bit. There is no way to verify what they are really doing and it seems just too easy for someone to throw the election with the right change agents working down at the old vote machine coding shop.

--

Regarding IDs, I certainly hope they are never required! Not all US citizens who are legally qualifed to vote are also allowed to get an ID. IDs are for the elite, not the disenfranchised. Let's not take even their right to vote away.

X. J. Scott
Wednesday, November 06, 2002

In New Hanover County, NC we use the scan-tron type ballot.  It was a fairly painless process.  Check in, get the ballot, bubble, scan, and go.  The ballot was a double sided 6" x 20" ballot.  You could even feed the ballot in the wrong way (back side up, or bottom in first, etc) with no problems.

apw
Wednesday, November 06, 2002

"Actually, Joe, real time results can be big disadvantage - people who plan and run elections tend to be concerned about information revealed influencing peoples' voting patterns. "

I agree with you Roger.  The GA SoS didn't start counting ballots and publishing results on the web site until after all the polls had closed.

Joe Paradise
Wednesday, November 06, 2002

"Can anyone name any advantages to touch screen type system?"

I think a big advantage is the ability to change or update the ballots at the last minute in unusual circumstances, like Lautenberg in New Jersey or Mondale in Minnesota

Joe Paradise
Wednesday, November 06, 2002

""Can anyone name any advantages to touch screen type system?"

I think a big advantage is the ability to change or update the ballots at the last minute"

You mean when the voter tries to press the "wrong" button it automatically swuitches the button text to represent the correct choice? :-)


Wednesday, November 06, 2002

Regarding ID requirements, or lack thereof, X.J. Scott says that a lot of people who are eligible to vote aren't able to get IDs. Is that really the case? Every U.S. state I'm aware of allows anyone (including non-citizens) to get photo IDs, but I'll be the first to admit my knowledge is not comprehensive.

I understand concerns about ID requirements preventing people from voting, but I also worry that the absence of ID requirements allows some, uh, creative misuse of the system. I wonder if there's a happy medium -- allowing people to vote provisionally without ID, subject to verification? I guess then you'd have to have a way to identify the potentially suspect ballots, though, which could be problematic.

John C.
Wednesday, November 06, 2002

One big advantage to touch-screen type ballots that I see is get feedback that your vote was actually registered.  When you press the button, it puts a checkmark there or something.  You can be sure that your vote was actually recognized.  With scantrons, you can never be sure... maybe your bubble was too small, or too light, and the machine won't pick it up.

I'm not sure we are actually ready for touch screen ballots, given that a large segment of the population is still not comfortable using computers.  I think their time will come though.

The security and quality of the software will certainly be an issue, though that issue does exist today in slightly different ways (trusting the machines, the people who make the "keys", the people who tally the votes, etc).  The software would certainly have to qualify under very strict government standards such as those in place for systems on aircraft

Actually now that I think about it, the "feedback" issue is the main problem I have with scantrons.  If you were able to run it through the machine and see instantly that all of your votes were recognized, I think that system is reasonable.  My understanding is that currently you will see if there are any gross mistakes that would discount the ballot, but you can not actually tell that all of your bubbles were recognized

Mike McNertney
Wednesday, November 06, 2002

As for IDs...

The "provisional voting" idea won't work.  There can be no way to associate a ballot to a person, or else the system is not really a secret ballot and is open to abuse.

As far as I know every state issues photo IDs... all you have to do is prove who you are via birth certificate and social security card, among other choices.

I'm not sure though that the lack of ID is a real problem.  You have to be registered to vote, and you can't register someone who doesn't exist or someone who can't vote.  Obviously there is the case of taking someone else's place at the polls, but that is not too different from making them vote a particular way via force or bribery, which would be possible regardless

Mike McNertney
Wednesday, November 06, 2002

With interactive computer systems,
"You can be sure that your vote was actually recognized."

How? Does it print a receipt at the end, which then gets deposited into a lockbox? Or does it say "You voted No", then you leave, and the 4-bit counter for No overflows from 16 to 0 while the 16 bit counter for Yes stays the same?

Note that the 40 year old mechanical systems have similar problems, though some produced paper tape.

In any case, you read the results; if the results are close you demand a recount, if they're not a few little errors don't matter.

mb
Wednesday, November 06, 2002

Here in Virginia, we have completely different booths from the ones described.

Our booths are surrounded with a cloth curtain, which hangs to approximately waist-height.  Nicely secure.

The actual interface is beautifully simple:  It's one big panel, with different voting choices printed on boxes scattered around the panel.  Each choice has a raised button next to it, and the button has a light on top of it.

When the booth is activated, the lights on all the buttons blink steadily (on for a second, off for a second, etc.).  When you press a button, the light on that button goes steady, and the lights on the buttons for the other choices in that section go out.

This makes it very easy to see which sections are incomplete, since they're the ones that are flashing, and they attract your attention a little more than then completed sections, which are steadily lit.  But it's still easy to see how you've voted, and since everything is on the same large panel, there's no need to scroll around from one page to another.

To complete your transaction, you have to press a big mechanical green button labelled "VOTE".  This button is set below the bottom of the panel, which I think encourages your mind to realize that you're moving out of the current context, which prompts you to review your work.

(I'll note that the lights are fairly small, about the size of the tip of a Christmas tree light.  So this isn't a panel full of giant flashing lights, like a control panel in a bad science fiction film.)

Brent P. Newhall
Wednesday, November 06, 2002

Regarding state ids, most state ids are available to aliens and non-citizens and do not mention citizenship status. Thus they make a poor proof of citizenship. In fact, aliuens in many states are not suxxject to the same requirements that US citizens are and have special exceptions that make it considerably easier for them to get state ids than for a citizen to do so. For example, here in Tennessee you can not normally get an ID without some sort of paper documentation such as a birth certificate -- unless you state you are an alien -- and then it's OK.

I have done much volunteer work at homeless shelters and I can assure you that ID is not available to all homeless people, regardless of their citizenship and residential status. Theoretically perhaps yes but practically... no. A homeless man or woman might as well be a black man in the south facing a literacy test 120 years ago -- theoretically he is said to have the right to vote, but in practice he does not.

Constitutionally, no one can be denied the right to vote who is of the right age and a citizen and does not have felony convictions. Requiring ID would mean that many people who are supposed to be able to vote would not be able to.

--

Regarding my comments about the insecurity of touchscreen balloting, it seems I predicted the future -- yesterday's  election was rigged in some parts of the country:

http://www.sltrib.com/11062002/nation_w/14013.htm

> In Tarrant County, Texas, a computer programming error was skewing the picks of straight-ticket voters. The snafu came on the heels of complaints in Dallas County, where the Democratic Party filed suit after early voters complained that the machines had changed votes for Democrats into votes for Republicans.

> In Florida, polls opened on time, lines were manageable and most of the new touch-screen voting machines functioned properly. But, this being a Florida election, there were some missteps. About a half-dozen of Miami-Dade County's 700-plus precincts reported problems with voting machines, with some registering votes for Bush after voters selected McBride.

>One voter, Reginald Smith, said he thought Bush's name kept popping up on his screen because his forefinger was too big, so he tried to touch McBride's name with his pinky. It still didn't work. "I'd rather punch," Smith said. "I know who I punched for in 2000."

In my opinion, computer balloting without a paper trail must be permanently banned. It is anathema to democracy. There is far too great a potential for fraud, even fraud transparently disguised as 'programmer error'.

X. J. Scott
Wednesday, November 06, 2002

How does it tell you if your vote is recognized?  I dunno... maybe it puts a big checkmark in the box next to the person you selected.  Or the lights thing that someone mentioned.  The point is there is feedback, whereas there is none in a paper ballot system.

I agree that bugs are a potential problem.  But come on, we trust our very lives to software each and every day.  Do you really think we aren't capable of designing a computer balloting system that works?  I certainly wouldn't want such a system to go "live" before extensive testing, and I agree that at least at first there should also be a paper trail (maybe printing out a sheet with a barcode-like indicator of the votes, along with actual text indicating the votes so it can both be verified and if there is any question you can just count them by reading the text).  But that doesn't mean we should never have computerized balloting.

The thing is, there will always be a lack of precision with paper balloting.  Every time you count them, it comes up different.  Is that really a better system than computer balloting?  Should we trust elections with such small margins to a system with a significant margin of error?  In that sense, one advantage of computer balloting is that it can provide multiple ways of counting, by keeping track both electronically and on paper.

By the way, I'm not talking about people getting confused and voting for the wrong person.  On any well designed ballot that will be kept to a minimum, and there is really nothing that can be done to prevent it.  What I'm talking about is abiguity in what is interpreted as a vote.  In the paper ballot system, the abiguity is not resolved until the ballot is counted, long after the voter has any chance to correct the problem.  With a computer ballot, the ambiguity is presented immediately to the voter, who can then correct the problem.  Say I pressed on the edge of the button for a candidate... Did that register or not?  Well, I can look at it and see if the computer drew a big checkmark there, and if not press again to correct the situation.

I just don't see how a paper ballot can get around those two big problems.  If you come up with a way, I'd be all for it and then I don't really see a need for computer balloting.

Mike McNertney
Thursday, November 07, 2002

One more thing... the last post mentioned some problems that came up with computer ballots... Were these things even tested?  I mean come on, those sound like incredibly basic problems.

Obviously it is bad that there were problems.  However, I don't think we should take that as a sign that we should never use computer ballots.  How many bugs, crashes, etc can you think up that you've seen in software in the last week?  Probably dozens, if you use any commercial software on a regular basis.  Does that mean we should not trust software to run our airplanes, cars, etc?  No, obviously not, because that software is held up to standards a thousand times stricter than commercial software, and voting software should be held to the same type of strict standard.

Mike McNertney
Thursday, November 07, 2002

The difference between flying an airplane and voting is cheating. If my control software decides to roll over a counter at 16 bits, the rocket explodes. If my voting software does the same thing, my prefered candidate wins.

The financial industry is much more relveant than the control software industry. The the same incentive to cheat is present. The advantage they have is that it's OK to tie transactions to people, which isn't necessarily the case with a voting machine.

If all voting machine makers were ethical and uninfluencable, we'd be a long way along the trust process. If they voting machine makers don't even allow inspection of the software they wrote.

You still always need a paper receipt. Preferably one you can read right there. Would you accept a scantron machine with a display the votes you scanned, and allowed you to hit a 'reject' or 'accept' button? That raises other issues, but think about it. It's really good to have a human-generated piece of paper in the ballot box, much harder to forge than a receipt.

mb
Thursday, November 07, 2002

I have wondered why NCS or Scantron did not make more inroads for ensuring that votes really do get counted.  Okay, I realize that errors do occur, but having an electronically scannable + paper ballot seems ideal.  Online voting could even "create" a Scantron style answer sheet.  Whether online or by paper, why not add a step where the input is shown on the screen and the voter verifies that these are the correct "answers" that were intended.  Then half circles or misprints could be corrected before entering the system.

MJTaylor
Monday, March 08, 2004

*  Recent Topics

*  Fog Creek Home