Charles Stewart III (MIT), Voting Machines: Friend or Foe? Here is the abstract:

Over ninety-nine percent of all ballots cast in the 2020 presidential election were counted by a computer. It is hard to imagine that underlying such a ubiquitous use of technology lurks serious questions about the safety and security of elections conducted with the assistance of voting machines. And yet those questions persist.

All this controversy boils down to two questions: first, do voting machines make it more likely that the choices of voters will be recorded correctly, and, second, should we trust the machines to count the votes correctly?

I argue that the general answers to these questions are “yes” and “no,” although the full answers are much more nuanced than that. Computer technologies extend the capabilities of humans in voting, as they do in most other aspects of modern life. Expectations of voters and policymakers about the electoral system are such that voting in America would be impossible without the assistance of computerized voting machines. At the same time, the nature of computer programming is such that one cannot simply assume that they perform as expected. Election results arrived at using voting machines must be independently verified. Standard operating procedures in place strive to provide such verification. However, in most states, such verification relies on antiquated approaches and incomplete measures that demand improvement.

This paper is organized in three sections. Section 1 is an overview of the voting technology ecosystem and how it is used to administer elections. The ubiquity of voting technology in election administration raises the obvious question of “why?” which is the focus of Section 2. Section 3 provides a conclusion that reflects a consensus among academic experts and most election administrators—that elections should not be conducted in a way that relies on the proper functioning of any technology, even pencil and paper. Rather, the security of elections depends on the ability to reconstruct independently the critical processes undertaken in an election and provide evidence that the results should be trusted within a previously agreed-upon margin of error.