You may have seen the media coverage
of the Australian National Audit Office’s report into the AEC’s conduct of the 2016 election.
Eyebrow raising to say the least.
Back in 2016, I was approached by Vanessa Teague
(we’d been previously introduced online by Asher Wolf) to help
out a team who were looking to conduct a Bayesian Audit of the Australian Senate Vote as cast at that year’s
That Bayesian audit required an implementation of the Australian senate count, and
I had one handy.
It turned out to be a really cool little project, and produced this pre-print paper. My main contribution was
turning dividebatur, Python code for running the senate count,
into a library that could be imported by the audit implementation. There was also a fair bit of work speeding
the code up.
The result was this pre-print paper: B. Chilingirian, Z. Perumal, R. L. Rivest, G. Bowland, A. Conway, P. B. Stark, M. Blom, C. Culnane, and V. Teague. Auditing australian senate ballots. arXiv preprint arXiv:1610.00127, 2016.
If you grab the PDF, you’ll find links to all the software that was produced, and you shouldn’t have any trouble running the audit yourself.
Having been involved in that project, my biggest take-away is this: there are ways that we can ensure the
integrity of the result of our elections, so that we can be confident in the result. Those things aren’t
unknown - researchers have already found effective means - and they’re not expensive, either. They do
take some work and commitment, and they require our electoral agencies to be open to the risk of failing
It’s vital that something is done to ensure public confidence in the ballot box, or we’re creating another
reason for public confidence in democracy itself to be diminished.
Image credit the banner image is derived from Victorian-senate-paper-folded by Hshook @ Wikimedia Commons.