Perhaps the problems I been discussing regarding our fallible perception of numerical information just relate to situations, like the examples I’ve used so far, involving statistics.
As Daniel Kahneman, who won a Nobel Prize for creating the field of behavioral economics points out, humans do not have great instincts when it comes to navigating statistical information and questions.
For example, if I asked people if they would rather take a bet with a 5% chance of winning $100 and a 95% chance of losing $5, or buying a $5 lottery ticket with a 1 in 20 chance of winning $100, most people would go for the lottery ticket.
But if you stop and think for a moment, if you win the first bet, you get to keep full $100, but if you play the lottery you’re out the $5 you paid for your ticket, win or lose, meaning the most you can win is $95. In this case, our instincts (specifically, our instinctive desire to avoid a wager with such a high chance of losing) overwhelmed our ability to reason statistically.
Even if our instincts were better than they are, statistics – like the one used to determine flu deaths mentioned last time, or stats crunched to determine who is up and who is down in the latest presidential poll – can be complex, requiring a grasp of mathematics and some sort of statistical training to understand. But what about much simpler mathematical operations, such as counting? Certainly something as ridiculously simple as counting things can’t get us into the kinds of trouble I’ve been describing involving the messy imperfections of numbers representing real things in the real world.
If you believe that, perhaps you have already forgotten the chaos of this year’s Iowa caucuses (which is easy to do, given everything that has happened in the world since then).
You would think that counting stuff – the simplest of all mathematical operations – might be free from the ambiguity I have been describing. But if you ever lost track while trying to count just a few dozen similar items (like counting to figure out if a deck of cards is complete), you’ll realize that the human mind’s tendency to wander and miss things (like that forgotten joker) makes even simple counting tasks susceptible to imprecision.
Now if you scale a counting task up to include not just you but thousands of people counting millions of different things in different places at different times, you can begin to see the number of potential sources for error. Given that every election ever held involves these sorts of distributed and complex counting tasks, there is likely to have been error (or at least a certain amount of uncertainty) with every vote bigger than a local election involving just a few hundred identical paper ballots that can be counted and recounted to ensure accuracy. Fortunately, the margin of victory in most elections is large enough to swamp such uncertainty. But when it doesn’t: Boom!
The most dramatic example of this was the 2000 Presidential election in which the aggregated choice of over a hundred million Americans came down to votes made by a few thousand people who struggled with a confusing ballot in Florida. This confusion was compounded by how a few dozen election officials, lawyers, and judges decided how those Florida votes would be counted. Throughout the entire chaotic affair, no one was willing to admit that in a tight-margin situation involving confusion, arbitrary decisions, and potential manipulation (on top of regular human errors that happen when people count things), “the answer” – i.e., a number that accurately reflects what the electorate wanted – might not exist.
If people had been willing to live with such uncertainty, options such as a do-over election might have been acceptable. But because of our faith that “the truth” in the form of a perfect number existed out there somewhere, we ended up living in chaos for weeks as votes were counted and recounted in hope of discovering a clean, accurate number in a messy, imprecise world.
Now the number of things one needed to count during the Iowa Caucuses was far smaller than the number of votes cast in a national election. But the things counted in Iowa were not pieces of paper but people who caucused for different candidates and were allowed to (or, in some cases, required to) change their minds if their first choice did not rise above a certain threshold.
Because caucusing involves tracking something more dynamic and complex than paper ballots or readouts from voting machines, we should not have been the least bit surprised that no one was able to answer the only question we were interested in (i.e., who won Iowa) before we went to bed on caucus night.
Unlike other years, this year’s caucus officials had to determine not just a final count but counts at every stage of the caucusing process to support transparency. In theory, a new app developed to help those officials manage this information should have simplified things. But like so many situations, adding technology to a complex human process simply introduced new sources of confusion (especially among vote-counters who were not used to using such tools).
So we had to wait for results that turned out to not make a difference in the end since the man who came in fourth in Iowa – Joe Biden – eventually won the nomination. All this confusion caused many to question the continuing role of Iowa in selecting presidential nominees. But it did not seem to shake people’s faith in the notion that numerical truth must exist, if we can only keep track of the numbers and keep counting and recounting until we get the “accurate” result.
Our Messy Election
Keep in mind that in a “normal” presidential election, we select candidates through many different processes practiced in each of the fifty states and the District of Columbia (machines, punch cards, write-in ballots of different formats), and that states also allow for early voting and mail-in ballots, options that are going to be used far more in 2020 than ever before due to concerns over Coronavirus.
I just sent in my mail-in ballot for a state primary, a process involving filling out the same form I would have used in the voting booth, sealing it in an envelope, writing my signature on the outside of that envelope, and then stuffing the whole thing into a different envelope and dropping it in the mailbox near City Hall. What could go wrong if millions of people choose this option during this year’s highly contentious election?
Quite a bit, apparently, if New York’s 2020 primary, in which more than 20% of mail-in ballots were disqualified for a variety of reasons, is any indication. And even if every voter and mail sorter does everything right, what if millions of ballots that are mailed in (versus cast on Election Day) can’t all be counted by the time the polls close?
The ability of the US postal service to deliver mail-in ballots and get them to the right place to be counted has been the subject of considerable, heated debate. But as Florida in 2000 and Iowa in 2020 (and other elections before, between and since where the margin of victory was within the margin of error) demonstrate, what we should really be worrying about is what is going to happen after those ballots have been collected.
If this is destined to be the highest turnout election in decades, we are going to have a lot of counting of a lot of different things filled in by millions of fallible human beings, then counted by thousands of other fallible humans to look forward to.
Hopefully the margin of victory for whoever wins will be sizable enough not only to overwhelm doubt over who is victorious, but also limited the credibility of accusations that the election was stolen. But if we want to prevent a patient approach to the task of vote counting from turning into yet another 2020 crisis, perhaps we should begin to get comfortable with the possibility that we may not have an instantaneous answer to the question of who won and who lost once polls close on election night.