[LINK] AEC faces backlash over vote counting ???black box???

[Chris Maltby]

The key (as David says) is auditability of the process.

The new system of counting is effectively fully electronic, but
scrutineers can watch the data entry process of a clerk which will
be compared with a successful OCR of the scan. The scrutineer can
also request and be shown the physical paper at any time.  I would
hope that scrutineers do this often enough to make the margin of
error for failure to detect a significant tampering with the recorded
data low enough to give assurance to the results.

The other audit capability is the (incomplete) counts of senate
first preferences by group that was conducted manually in polling
booths on election night. This data is available for statistical
comparison with the booth-by-booth final vote data and that would
also show up any significant favouritism in the data entry process.

Finally, it's worth noting that the old senate counting system of
a manual sort into 1-above-the-line groups and others, done in the
divisional counting centres, is subject to a reasonably liklihood
of manual errors. The voided WA senate election of 2013 showed that
a significant number of above-the-line votes had been mis-sorted
at this stage and incorrect counts were then entered. Some were
also famously misplaced. The data entry process for below the line
votes was more accurate.

The new counting system will see all papers subject to dual data
entry, with at least one human operator and two if the OCR doesn't
deliver a high enough confidence result. A mismatch will see an
additional data entry step for the paper. If the process has the
integrity we hope for, the final data set is likely to be more
accurate than the old system.

Finally, it's worth noting that the changes to formality rules
should mean that there are a lot fewer ballots with all the
below-the-line boxes filled, which are the slowest to enter and
the ones most prone to voter and data entry error. It may be
that the overall data entry time won't be all that much longer
than in the past, as the vast bulk of papers will have just the
minimum 6 numbers above the line, with fewer using the below-the-
line option, and those who do probably numbering fewer boxes.

The data entry software identifies all the non-printed marks on the
paper and presents them one-by-one to the clerk, so a 6 or 12 mark
paper is pretty quick to enter as distinct from the former process
of entering the contents of every one of the 150+ below-the-line
boxes.

Chris

PS You won't hear any argument from me against release of the
   actual vote counting software's source code (and of course
   the actual entered vote data set as well).