Here’s what we know about what happened Monday at the Iowa presidential caucuses. The Iowa Democratic Party hired Shadow Inc., a startup company, that built an app that clearly hadn’t been tested well enough before it was deployed in the real world.
Nevada had planned to use the app, but has now said it won’t use it in its caucuses later this month. Sure, that app choked, but long term, this is a problem about trust when it comes to technology in elections. It also raises the question, what parts of the voting process do need tech innovation?
I spoke with Charles Stewart, a professor of political science at the Massachusetts Institute of Technology and a member of the Caltech/MIT Voting Technology Project, and asked him that question. The following is an edited transcript of our conversation.
Charles Stewart: The best use of technology in elections is to do the things that human beings are bad at doing. Human beings are very bad at doing tedious things, and the most tedious of those things in elections is counting the ballots. The best use of tech, in my view, starts with a really well-designed system that uses computers appropriately to record and tabulate the votes, but then has paper systems that human beings can verify the vote — make sure that the computer recorded their vote properly — and that you can go back to the paper to verify the count if you have to, either in a recount or because you’re doing an audit.
Molly Wood: How big a market is there for what you have described and for election tech overall?
Stewart: There’s a market for voting systems. There are, I like to say, two and a half companies that produce these systems, meaning there’s two large companies and then there’s a number of smaller companies that produce tabulators, produce screens for those who want computerized screens. But it’s a very small market. People in the elections have been hoping to expand that market, either by making it easier for new companies to enter or to find ways to take off the shelf parts and to create voting systems from iPads and HP printers, etc.
Wood: Let’s talk about the competition piece, because it sounds like some big tech companies have actually been in this market but then gotten out, right?
Stewart: Right. The best and most famous story is that IBM was actually responsible for the proliferation of the punch card technology in the early 1960s. In the 1960s, an internal report from IBM concluded that one day, inevitably, something would go wrong, and Big Blue would be left red in the face. So the decision was made to sell that part of the business. Right? They were lucky, they were not the company that was blamed for hanging chad in Florida in 2000. These systems are in the spotlight, they’re highly visible, and we know that there will be mistakes. It looks like those mistakes are really bad for business. It shifts the provision of technology from the really big providers to smaller and smaller providers.
Wood: How big a deal is the education part of it when it comes to even the Democratic committee choosing a vendor? We have to accept the possibility that the people involved here are just not that tech savvy, maybe at every level. Is there a solution for that?
Stewart: I would bet on another scenario, which is that people are tech savvy enough to think about the typical tech applications that you might develop. They think, “Well, how hard can it be to create an app that takes some numbers that you type in over here and push them over to a data server over there?” Technologically, it’s not necessarily a big problem. There are other issues, though, having to do with the scrutiny that the system is facing and the consequences for getting things wrong and the ability to vet the software and to try to break it. Basically, make sure that it’s bulletproof. I’ve seen a number of instances where people have just radically underestimated the complexity of building an election system of any sort, thinking that it can’t be any harder or more complex than other things that they built — and they’re often wrong.
Wood: When something like this happens, it sounds like on the one hand you’re arguing that there is a need for better standards and communication, possibly more competition. But what is the cost of an event like this to voters’ trust in new technology?
Stewart: The cost has to be pretty high. First of all, it’s incumbent upon people running elections or caucuses to run them in a way that it looks like it’s free and fair. Most people who are not involved in this business don’t recognize how chaotic it really can be. You really have to do your due diligence to make sure that it looks like people know what they’re doing. Whenever it looks like people don’t know what they’re doing, then it’s not going to be good. My expectation is that as we move on to other states, I’m hoping that Iowa will fade in people’s memories and this hit to democracy won’t be as tragic as some people are worrying it will be.
Related links: More insight from Molly Wood
The Iowa Democratic Party actually doubled down on tech in recent years as a way to clean up elections and make them more secure and reliable. But ultimately, this is less of a problem of technology itself and more a problem of how to implement new tech. This is something cybersecurity experts have been talking about in general, but also around elections for years.
NPR and the Wall Street Journal reported on the Iowa Democrats’ decision to switch to this smartphone app. They spoke with security experts who were raising concerns in mid-January about the little transparency there was around this app, as well as how it was tested, how much it had been tested and where it came from.
If you want a deep dive on the economics of voting technology, there’s a Wharton School report from 2016 that looks at the market for election tech, the vendors, the security concerns and the regulation around it all.
Also, what Charles Stewart told me is that while there are voluntary federal standards that dictate what should go in a voting machine, there aren’t really standards for all these other technologies that could become part of the election process, whether it’s vote tabulating, reporting or data gathering. Clearly, by now, there probably should be an update.
This is the final word on the voting software topic still valid today, a comic by XKCD, which has been right about everything.