,
8 tweets,
2 min read
Read on Twitter
1/ More on the science of voter verification of ballots: Those who think voters can improve their error detection effectiveness are flying in the face of established science and the consolidation of error studies across many disciplines that took place in the 1970s-80s.
2/A major reference is James Reason's 1991 "Human Error"
cambridge.org/core/books/hum…
cambridge.org/core/books/hum…
3/Here are the take-aways that IMO make it difficult for voter-verification. Studies show error detection is much harder than the original task. That is, while diligent people can be trained to minimize errors in cognitive tasks, error detection rates hover around 50-80 percent
4/There is a unified model of cognition that explains what factors affect error detection. Rates do not vary very much from task to task. Proof reading errors for non-spelling are a good exemplar for tasks like ballot verification.
5/There is wide variation from individual to individual. Experience at the task matters but not by much. In tasks like ballot verification, a reasonable expectation is that base rates can be improved upon by only a few percent.
6/Error reporting/correcting seems to require the intervention of a third party.
7/Environmental factors unfavorable to the task of ballot verification have a pronounced negative effect on a person's innate ability to detect error (like in this Basili-Selby study of error detection in code reviews) 
