For some of the tables that this is equal to this, and for other tables that this is equal to this. That never reveals these sums until...That way, we know these are copies of each other except for the roads that are marked decoy or nothing is known about them.

This one corresponds to that one. Here, you see this one is equal to that one except for the ones marked decoy, and that one corresponds to that one in this case here. These are a little bit folded together to make it more efficient, but basically, we audit...

No one will ever know that you’re a decoy voter. That I didn’t include you in advance of the coin flip, because I never open these. I let people check. You see, this one...

Before the purple, I think I have to accept the request for decoy. Because then I know what numbers to put in the green when I publish it. I say, "Oh, you wanted to be a decoy voter? Great. I’ll write you in in the green."

Yellow is committed before the purple choice. That’s the coin flip thing. But the green, these yellows and green, you see how they’re mirrors of each other? Some of the yellows have the decoy. The decoy slots were built in advance.

The little bit of tricky business in here, you see the reason we break the voter number into one of the summands is because we want that to be random after this, and this stuff is committed.

That’s important because if number of the decoys could be, we don’t want to give them to wrong people. "Did you get ballot number 100? Did you vote it?" "Yes." "OK. Good. Then we’re done with that audit." Does that make sense?

In that final step, you would say, hey, you finally can find out the third summand. That’s the 222. Now, we know to go to the house of the voter number, 777, and then we can say, "Hey, did you in fact get ballot number 100?"

Then there’s this thing they sometimes call canvassing, two weeks after the vote to check everything. That’s this part here. Or you can do it all at once or in these three conventional stages.

We can do this in stages. There’s a preliminary audit which just locks in the recording of the ballots, and the printing fraud that we talked about, and there’s a second stage where we actually reveal the votes. It’s like reel elections because when we run reelections, there was like ...

Depending on which of the five categories of this is in, we have to open certain tape. If it’s in the first category, then we have to open these two tapes. We reveal those three columns. If it’s in the second category, then we reveal this column, and this column, ...

It’s actually interesting. Then, we do this division by the coin flipping of all of the spreadsheets into five categories. There’s like 50 spreadsheets in each category. We choose which one go...That’s by coin flips, we could not have known that in advance.

These are the nine stages of the election. After the voters log their ballots on the voter bulletin board, and the election officials then finally...I lied. We have to say, "Guys leave the room now, or we’ve got this other room where we’re going to write the blue columns. We’re ...

The coins tell us which tape to pull off. Now if you look at this chart, it’s very exact about which tape to pull off. Let me try that. Indicate that to you. You see these bars below here, those lines are...Let’s say this line here means that there’s the ...

Then we give out the ballots which we’ve written up, and they vote them, and then basically you just, "Come in, and write your vote codes on the wall." They all come in, they write their vote codes knowing that their ballot secrecy is protected. Then they sit there while ...

We write all these up on the wall, and then we put the black tape down each column, and then we invite the voters in, and say, "You’ll stay in here for any election. Watch that no one comes in, and changes any numbers under the tape."

There’s a lot of difference, what do you call it, separate group ways to add. You can have two numbers that have the same sum. There’s the full number. Those pairs, we randomize those pairs by like panning around a numbered one, to drag it from the other, and then ...

There’s another thing which is done which is those pairs, the columns at the end, the yellow and the blue pairs, the sum of adjacent entries in the yellow pair is the same for all the rows before the permutation.

Then for the consistencies, the algorithm for creating that is describing there basically all the stable are the same except they’re subject to random row permutations. The rows are reordered, but they’re kept together.

We would come in here at night, and write these on the wall with a marker, and then we’d cover each column with a piece of black tape. You and I, we did that all by flipping our own coins, or whatever. We’d just make this stuff up.

It says there are 256 such tables. Do you see those indices? Yes. It’s actually like 256 rows. These are generated numbers when we use computers, but spreadsheets which are...I think it’s, what is it, eight columns of these short text strings.

What we would is we would write those numbers there that are shown. Those are lists of numbers. You see that. There’s a bit of text in them. There’s ellipses because for every ballot, there’s a row. There’s 1,000 ballots so it’d be 1,000 rows, but you see, there’s also ...

Here’s a physical analogy for how you could it which is extremely close to the way it works. It’s essentially the same thing. Suppose let’s say you and I were going to run an election, and there were no computers or anything, and always plan to use a new computers ...

When you go to vote the ballot online as we call it, electronic bulletin-board, it’s pretty sophisticated the way we’ve done it in practice in the election.

It’s very clear. It’s not such a big problem, but there’s a finesse where you can get rid of that. I’ll just mention it to you briefly if it’s not shown here, or should we talk about how this system works?

That’s how we do it, but in a voter sense, it’s a little weird to get two. We did it where you put one on one side, one on the other. It’s livable.

That way, that’s pretty clear that if there’s what we call print fraud, then there’s a 50/50 chance that instance of it will be detectable by a voter having this half that doesn’t reflect how they voted at all. It’s hard evidence of malfeasance.

The one that you see on the left there, the 100a, and the one...No. Down on the left. The extreme left. You see, that’s a printed form. That white thing. You see how there’s two versions of the ballot. The idea is it says, "OK, voter, you choose one of ...

I’d really love it if you did understand that picture, and I want...I was hoping to explain it to you, but you see how you have a double-ballot there. It’s like it would be ballot say 100, and there’d be two forms of it.

That’s a technicality and if you read the random-sample voting paper you’ll see that we use what we call, and it’s confusing, but it is confusing because it confused a lot of people, double ballots. If you look at that last...

That would then, with some probability detect...Then we saw that it was wrong, because we could open, we didn’t, you could open that particular ballot then and there but probably we didn’t, we opened it at the end. You’ll understand later why we did it that way, but probably you ...

You act, of course, just like a voter and you just come in get your thing with the rest of the voters, but then once you get your ballot from them, you say, "I am here to audit and I am going to mark every question and then under Maryland ...

In any event, you should randomly, or at least rotate. They can be alphabetical but it’s sequentially computed, they should have done...The codes are, you only see the code that you mark, and so there is this malfeasance. In Takoma Park, we did this thing where someone as according to ...

If I knew that a certain candidate was likely to win and I didn’t like that, I could criminally print ballots that switched those...You don’t want the candidate, in my opinion...Everyone knows that you should randomize the candidate names. If you have too long list, you can’t do that obviously. ...

Look at where these are really technical about how these systems work, the one thing you might have not have noticed too much is the only funny thing about it is that when you have a physical paper ballot, there’s a malfeasance where the vote codes could be swapped relative ...

We had a sketch for that. I know you can be very technical. Sometimes, you probably are not using those skills these days in your day job...

...it does a little bit of auditing, but it chooses a random place to check. Then in aggregate the auditing would be done anyway by everyone.

Now, here’s like a technical hacker perspective on this, "It’s a certain amount of work to check everything," but actually, what we never wrote, but there was a German guy that got halfway through it, but what you really ideally want is a smartphone app that while you’re looking to ...

We have multiple people whove written their own independent auditing softwares and then the APIs are public and documented, people can write their own auditing software or they could run those auditing softwares from others.

Unbelievable. Either an election is publicly verifiable or it isn’t. These are publicly verifiable. In other words, you would present this as openly, I don’t know what you’d call it, transparently verifiable. [laughs] Anybody with a laptop or a smartphone can go and see that all the posted data really ...

Those elections we ran in Takoma Park, Maryland, 2009, 2011, the only publicly verifiable binding governmental regular elections that I know of even to date.

We offered any jurisdiction that would want to start adopting it, we’d give them all the software for free and all updates in perpetuity for free, but no one signed up. I don’t know what’s with that. If you’re running polling place voting, there’s no excuse for not using it, ...

Still, they were very happy with the overall system. I guess they were willing to take that second phase on faith somehow seeing at least their vote was...I don’t know exactly how each person thought about it, but the overall appreciation, satisfaction was much higher, and in fact, it costs ...

That’s something the survey showed distinctly people really appreciated. Very few people, of course, understood the proofs that we give of the verifiable proof that the tally does accurately reflect the votes that are encoded in those numbers.

You get a really neat result and people find that gratifying. Then the code is revealed. If you write those codes down, you can check later online that they’re posted. That’s like a FedEx tracking number for your vote. People found that really great, because it means that their vote ...

Of course, the voter satisfaction is a factor in voter turnout, right? This might be something you want to consider. Yeah, and it’s actually really fun, because when you fill it with the yellow highlighter, it just turns black in half a second and you see it. It’s as if ...

Then those responses were recorded, or summarized, or whatever. Then, we also had some focus groups with election workers and all this stuff. We published these results, and basically people loved it.

We did surveys of the voters when they leave the polling place. Every voter was asked to fill out a 10-question questionnaire and every third voter exactly was asked an open-ended question by a professional interviewer. Like, "What did you think of this voting? How did you like it? Can ...

Then we could just inkjet print the ballots. It’s a lot less expensive then the direct digital printing, which is used by the two companies in the United States, or something, that are allowed to print ballots.

You almost won’t see it anywhere in the world. It’s really a bit disappointing, because we found that it’s much less expensive than voting systems sold in the United States, because we use commercial off-the shelf office scanners instead of the special ones, which are more expensive.