Thread by @rad_atl

/1 Thanks to Brad for informed discussion. We’ve heard about other virtues of ballot verification that are not impacted by these results but lets focus: it is likely that most humans are unable to verify the correctness of ballots cast moments before.
bradblog.com/?p=12829

2/It’s not a matter of ballot design, verification logistics or other practical considerations. The most likely explanation is fundamental limit on human cognition. Limitation presents special problems for secure voting, including post election audits.

3/If confirmed by other experiments, it’s the limited nature of working (short term) memory that is at fault. That would present not only problems for verifiability but also for usability and accessibility for disabled voters.

4/Here’s why. Since GUIs (e.g., touchscreen ballot markers) present information a screenful at a time, voters can only retain a portion of the ballot in memory. A voter reviewing prior ballot choices before casting the ballot increases cognitive load.

5/Increased cognitive load means increased user input errors, memory errors, and decreased effectiveness. Instead of endorsing GUI based ballot markers, we should be conducting these kinds of experiments.

6/Despite the hopes of BMD advocates, it is not a slam dunk that these devices add any value and significant evidence that they detract from security, verifiability, accessibility, usability.

7/Before rushing ahead with untested advocacy of ballot marking devices, I recommend a (re)reading of Irving Langmuir’s 1953 talk on “pathological science.”
cs.princeton.edu/~ken/Langmuir/…

8/ Langmuir talked about the science of things that aren’t so, which, among other things, cautions against subjective effects, wishful thinking or threshold interactions. Is the “science of ballot verification” in that category?

9/Not my first verification rodeo:
cs.umd.edu/~gasarch/BLOGP…