November 23, 2000
RAY SUAREZ: Who knew counting could be so complicated -- an initial count, several recounts, machine counts, hand counts, dimples, chads and so on. While law and politics will determine the outcome, science may offer some insight. Here's where things stand. Nationwide, Al Gore received 50,133,912 votes -- George Bush, 49,805,216. That's a difference of 328,696 votes out of 103,772,392 cast. In Florida, the current official tally shows George Bush with 2,911,872 votes; Al Gore with 2,910,942 votes. A difference of 930 votes out of 5,960,712 cast. Joining me now is mathematician John Allen Paulos of Temple University. His books include "Innumeracy" and "A Mathematician Reads the Newspaper." And physicist Neil Gershenfeld, director of the "Things that Think" program at the MIT media lab. He's author of the book "When Things Start to Think."
John Allen Paulos, let's start with you. When a mathematician looks at a spread, like this one, of fewer than a thousand votes out of about 6 million cast, what do you see?
JOHN ALLEN PAULOS: The results are indistinguishable from chance. Were one to flip a coin six million times, the difference between the number of heads and the number of tails would be comparable. It's less than, to use a technical term, one standard deviation from 50-50. So it is essentially a tie. The interesting thing about the race, I think, is that the imponderables are so much greater than the margin of victory, the margin of error is so much greater than the margin of victory - 20,000 votes in Duval County, 15,000 in Seminole County, the 3,000 butterfly ballots. These numbers swamp the 900-vote difference or the 700-vote difference between the candidates. And were there to be another recount, the numbers would be probably considerably different. Gore might be ahead if we recounted the whole state. Bush might be even further ahead.
RAY SUAREZ: But we don't do ties in American politics. I mean, you're looking at this through the eyes of someone who works with numbers, but we don't have a system that sort of embraces two candidates getting the same number of votes.
JOHN ALLEN PAULOS: That's true. It's a problem. We need something. Whether we resort to a randomizing device, a poker hand or a coin flip, a commemorative Bush-Gore coin flip in the rotunda in the capital building in Tallahassee or not, something has to be done -- perhaps a sudden-death debate on the Lehrer NewsHour. If Gore sighs before Bush smirks, he loses; if Bush smirks before Gore sighs, he loses. But some sort of randomizing device or split the electors. But, short of that it's not clear where we're going. I mean, any result is going to have a large degree of arbitrariness in it.
RAY SUAREZ: Neil Gershenfeld, is there such a thing as a technological fix for this problem?
NEIL GERSHENFELD: There certainly is, but understand this isn't a moon shot. This is really an excusably archaic technology. Punch cards were invented in 1890 and thrown out in 1960 by everybody but the politicians. And, in fact, it's interesting to go back to see where chads came from. In the '40s and '50s, there was a keyboard invented by a Mr. Chadless to punch cards without leaving a tab that would gum up the main frame. And so what was left was called a chad. And those were shutting down main frames that were causing problems in the '40s and '50s. That's why the punch cards were gotten rid of it. Very modest technology could do an accurate count of what people intended to do.
RAY SUAREZ: In the political debate surrounding the count, we've set up this continuum where at one end there are those who say that bringing people back into the process would take out some of the randomness and bring back some of the accuracy and at the other end of this continuum, there are people saying, no, no, putting machines back at the center of this question will assure accuracy of the count. What do you think of that continuum?
NEIL GERSHENFELD: The important thing to understand is there are deep questions about the role of technology in our society, and they're not posed by this election. This election used ancient technology. There's just a few modest steps. You need to have feedback to confirm from the machine, did the person do what they intended to do? And you need to securely distribute the data on a network to get it where you want it to go to. Those aren't deep questions. With those you can get an accurate count. So this discussion of machines doesn't distinguish between bad machines that don't work well, which is what was used in Florida, and modest technology that is within a few decades of the present that would work well that would address those problems.
RAY SUAREZ: But perhaps it does to the extent that some people feel more faith when they see you put a stack of cards into a sorting machine that counts them, there is a feeling that some people bring to that action that has... they have more faith in what the machine comes up with than people sitting around a table counting by hand.
NEIL GERSHENFELD: Good technology, you don't think of technology. My pen is technology but it works so well you don't think of it as a machine. And that's how voting should be. The machines you see chunking away might be satisfying because they make a big noise and people stand around them. But for decades we've just known they're not reliable machines. A good voting machine easily, like say an ATM, would count what you intended to do, and this discussion would go away and you wouldn't ask, should we use the machine? It just works so well that you stop noticing it as a machine.
RAY SUAREZ: John Allen Paulos, let me go back to you on this question of faith. A lot of the integrity of our system comes from the credibility that we as an electorate bring to the results. They are credible because we believe them and we believe them because they're credible, in this great big circle. But you mentioned before that when the result is smaller than the margin of error, we start running into problems.
JOHN ALLEN PAULOS: Right. I think to some extent, you were talking about old technology, I mean I'm being hyperbolic here but measuring the difference between Mr. Bush and Mr. Gore's totals is somewhat akin to measuring the length of a bacteria with a yardstick. The system in Florida at present isn't sufficiently precise to register such fine discriminations. I mean, there are things that can be done as Neil just mentioned. I mean, one thing is besides it confirming the voter's intent, a machine could make it impossible to vote for two candidates for President, let's say. The 20,000 votes that are under dispute, they're not under dispute but that were thrown out in heavily Democratic Duval County, for example, because people voted for two candidates for President, there were a few on the next page and people thought they were a different office, all of those votes would not be cast because the machine would say, "you can't do this." So, you'd have 20,000 more votes in this case for Mr. Gore. So, comparably to the problem in Seminole County with absentee ballots, the election officials work them out. If they're thrown out, once again, Mr. Gore wins, that problem wouldn't exist had we an adequate technology, nor the problem Neil just mentioned. People voting for Mr. Buchanan by mistake because the machine would say, you have registered a vote for Mr. Buchanan -- is that what you intended to do? Everyone is familiar-- not everyone-- but most people are familiar with buying items on Amazon, for example, and before you submit your final order, it says are you sure that this is the book you ordered and you press yes or submit and you get the book you want instead of some random book that you accidentally ordered because you mistyped a letter.
RAY SUAREZ: But there we go, back to Neil Gershenfeld. We're talking about visibility and faith in the result. Computers may be more accurate than punch cards, but there's something unseen about the way it works. What may be satisfying in some people's view is that a punch card, you can see what you've done, somebody else can see what you've done. And then a week from now, you can look at that same card and see what you've done, but you can't necessarily do it with a computer or at least not in a visible, obvious way.
NEIL GERSHENFELD: This focus on the punch cards really is just an awkward adolescence in how we use information technology. Ancient machines like, say, a piano is technology and you trust it. You don't think about how it works and the Internet is technology. You don't think about how it works. And you trust it. This punch card, we focus on do we trust it so much because, in fact, it is reliable? There's just decades of experience that shows, as a device, it doesn't work well. What I really want to underscore is there's very modest technical steps that give us machines that are so trust worthy you don't really have to think about it. You don't have to wonder when you close a light switch, will the light turn on? Voting very much easily can be like that with these small steps.
RAY SUAREZ: John Gershenfeld, Neil Allen Paulos, thanks to you both.