1/While I appreciate @benadida constant references to the shortcomings of hand marked ballot, it’s blindingly obvious that there’s no equivalence to machine marked ballots.
2/By whatever mechanism mistabulation occurs, RLAs are designed to confirm correctly reported outcomes.
3/I still do not really understand the argument about mistabulating bubble marks. I’m sure it happens, but evidently it happens so rarely that no one in the business of establishing benchmarks for high volume op scan devices has bothered to document it.
4/I want to deal first with the underlying problem with the human-readable ballot or ballot summary. In the HMPB case, auditors can fall back on what can be discerned about voter intent because they can hold in their hands the best available record of intent.
5/In the case of machine marked ballots, the durable record of voter intent is lost and is replaced by a machine-generated proxy that may or may not be fraudulent.
6/ If the voter never looks at the ballot (which happens at least half the time), there is no reason to believe that the proxy represents intent.
7/Even if the voter does look, error rates are sufficiently high as to be indistinguishable from the case in which the voter doesn’t even try to verify. That happens at least half the time too.
8/ In other words, when the auditor randomly selects a machine-marked ballot and adds it to the confirming instances of the outcome, the ballot cast by the voter might just as well have been a confirming instance of a different outcome. And the auditor has no way to tell.
9/You can audit only the human readable ballot you have, not the one you should have had.
10/Next, adding bar codes (as we point out in dx.doi.org/10.2139/ssrn.3…) does not by itself cause the audit to be less meaningful. Let’s say the human-readable BMD output and the barcode don’t match, That would surely be caught in an audit
11/However, there is no guarantee that both aren’t fraudulent. This ambiguity can be exploited by an attacker.
12/I admit to being skeptical that the "content" of a bar/QR code is a well-defined concept. Codes can embed nonprinting escape characters. Codes that match character-for-character human readable ballots can still pass control signals to malware effectively poisoning the codes.
13/Let me end by re-opening a can of worms first opened by @philipbstark All this passion about BMDs (including VSAP) has to be viewed in context. Who is supporting/not supporting and why.
14/It is fair game to ask (as Philip did) posters to disclose anything that might affect their independence. I have no stake in the outcome of this debate beyond my interest as a citizen.
15/@threadreaderapp unroll
