I want to try to clarify a lot of the furor I've been seeing here and elsewhere, so I'm gonna take a minute to do that here. This thread comes in N parts:
-
-
Prikaži ovu nit
-
Part 1: Thanks. There are so many people to thank for helping put this work together. First and foremost, my advisor,
@jhalderm for supporting my development as a researcher, for working through the study with me, and polishing the heck out of our final draft.Prikaži ovu nit -
My other coauthors were also absolutely outstanding.
@allismcdon,@henryjmeng, Jensen Hwa,@thenakulbajaj, and@KevinMChang all really stepped up.Prikaži ovu nit -
We were right in the middle of this study when my mom died, and I would in no way have been able to get this done in the way we did without each of these wonderful people. Last year was easily the worst year of my life, but these folks helped me at least salvage some good.
Prikaži ovu nit -
I'd also like to thank the Ann Arbor District Library
@aadl.@henryjmeng and I walked in expecting to be told that they didn't want their patrons bothered by a bunch of annoying CS researchers, and instead we were welcomed with open arms.Prikaži ovu nit -
Seriously,
@ulotrichous, Jackie Fleischer-Best, and everyone else@aadl went
to
bat
for this study, using library resources to help us get recruitment. I can't express how freakin' awesome that is. Talk to your libraries, y'all.Prikaži ovu nit -
I'd also like to thank
@MikeByrneHFHCI and Phil Kortum (who I'm now mortified to realize I forgot to thank in the paper's acknowledgements) for talking to me in the early phases of the study to help me get my head around it.Prikaži ovu nit -
(Also shoutout to Phil and Claudia Acemyan for collaborating on an earlier work and teaching me how to pretend I know how to run user studies https://mbernhard.com/papers/https-usability.pdf …)
Prikaži ovu nit -
I also want to thank
@mcranechilders,@benadida,@danwallach@philipbstark,@TypeMRT,@thejoshpit,@aleksessex,@JoeBeOne,@kskoglund, and many many others for listening to me incessantly talk about this study for months.Prikaži ovu nit -
I really hope I haven't neglected anyone, but if I have please yell at me.
Prikaži ovu nit -
Part 2: Old Man Yells At Cloud. I've been pretty shocked by the attention that this study has gotten.
Prikaži ovu nit -
I knew that we were studying a fairly contentious topic, and I knew that there would be much twittering about it, but I was really, really hoping that this would cool the discussion, not inflame it.
Prikaži ovu nit -
To that end, I must take issue with the way the study has been characterized by
@ericgeller at@politico and@Joseph_Marks_ at@washingtonpost in particular.Prikaži ovu nit -
That the study was given over to clickbait is not surprising to me, but I am disappointed in two otherwise excellent journalists. I am the lead author on the paper, and I was not contacted by either of them for comment.
Prikaži ovu nit -
Both stories brush aside what I think are _really_ significant findings of our paper: BMDs are insecure, but there may be ways to secure them. Both stories seemed designed to inflame tensions among researchers and activists. Both are missing crucial context.
Prikaži ovu nit -
This is bad science journalism. Alarmism like this has contributed to us ignoring scientists on climate change for decades and to mistrust medical professionals whose work contributes to an ongoing literature of understanding what makes us sick and what doesn't.
Prikaži ovu nit -
I'm not here for sour grapes (no press is bad press after all), I just want to call a spade a spade. If you're interested in what better coverage of this paper looks like, I was deeply impressed with the way
@AJVicens wrote up our study (https://www.motherjones.com/politics/2020/01/ballot-marking-devices-election-security/ …)Prikaži ovu nit -
There is thoroughness, context, and engagement in the science.
Prikaži ovu nit -
Part 3: Science. Science is an ongoing process. Our paper is one of many that help fill in the blanks on BMDs, both existing and work to come. Our study looked at one population in one place on particular model of voting machine.
Prikaži ovu nit -
We found real evidence in our population that BMDs are not secure. Were I making a purchasing decision, I would not buy an all-BMD solution. But this is one study, and science takes many studies before we can make strong, generalizable claims.
Prikaži ovu nit -
We also found that there may be paths forward to better secure BMDs. Emphasis on the "may". Again, this is one study. Replication matters.
Prikaži ovu nit -
I know I'm sounding kind of defeatist of my own work, but anyone who tells you they know anything for sure based on one scientific study is either suffering from confirmation bias or selling snake oil.
Prikaži ovu nit -
Part 4: Takeaways. Yes, science is an imperfect and ever-ongoing process. Yes, ours is one drop of water in an ocean of knowledge to be gained about ballot marking devices. But there are a few things I feel confident in saying after our study.
Prikaži ovu nit -
BMDs on their own do not appear secure enough to be used safely in an election context when they are the sole means of voting. This assertion is based on our work, but also prior and contemporaneous work by Selker et al., Acemyan et al., Green et al., and others.
Prikaži ovu nit -
However, an important caveat to this is that we don't really know what else is secure. Hand-marked paper seems to intuitively provide a verifiable paper trail. However, this has not been confirmed empirically.
Prikaži ovu nit -
What happens if the printer drops a race or candidate off a random 10% of paper ballots? Do people notice? We don't know.
Prikaži ovu nit -
So when you hear that BMDs are unverifiable, please understand that it does not mean that hand-marked paper *is* verifiable. It probably is, but we haven't established that.
Prikaži ovu nit -
Secondly, we found evidence that BMDs may not be as unverifiable as previously thought. Participants were three times as likely to catch changes in races at the top of the ticket in our _non-intervention_ experiments.
Prikaži ovu nit -
This indicates that an adversary desiring to change a high profile race has a higher chance of being caught.
Prikaži ovu nit -
We also found some fairly easy to implement strategies that had a dramatic impact on verification.
Prikaži ovu nit - Još 12 drugih odgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.