Useful Links: www.tomw.net.au/links

IT issues on 666 ABC Canberra Drive with Keri Phillips each Monday at 5:50pm

Tom Worthington With Tom Worthington FACS, Visiting Fellow, Department of Computer Science, Australian National University

Electronic Voting in the ACT, 22 October 2001

The Australian Capital Territory (ACT) was having a local election on 20 October. This was the first use of a locally developed electronic voting system.

The ACT Legislative Assembly has passed legislation that will allow computer voting and vote counting for the next ACT election in October 2001. The Commission advertised a tender for software systems in December 2000 and received 7 tenders. The successful tenderer, Software Improvements, was announced on 19 April 2001. The Electoral Commissioner has set up a reference group, consisting of representatives from parties, MLAs and special interest groups, to provide him with advice on the introduction of electronic voting and vote counting in the ACT.

Electronic voting at the next ACT Legislative Assembly election, Australian Capital Territory Electoral Commission, 10 July 2001.

The system, called EVACS, is in two parts:

The front-end is used in polling places to directly record votes, and the back-end handles the data entry and counting of votes. The front-end (polling place) system The system deployed in each polling place where electronic voting is available consists of a number (probably 10) of "booth" machines, which are used by individual voters to cast their vote. These are connected via a network to a single "ballot box" machine which acts as a server for the polling place, recording votes as they are cast and providing information to the booth machines when necessary...

The back-end (tally room) system The back end system will consist of a number of data entry workstations and the counting and control machine. The data entry workstations are used by data entry operators to enter manually cast ballots. These are recorded on the counting machine, which will also receive (via removable media) the electronic ballots. Once the ballots are received/entered it will then perform a Hare-Clark scrutiny to produce the results of the election, and also reports similar to those presently produced during the manual counting process. The counting machine will also be used to perform some of the administrative functions needed for the system, such as preparing the candidate and language information, which must be distributed to the polling places, and generating the barcodes which are used for authentication ...

Electronic voting at the next ACT Legislative Assembly election, Australian Capital Territory Electoral Commission, 10 July 2001.

As I would be away on election day I decided to use the electronic system for a pre-poll vote. As it might be of interest to others in the IT inditsry I wrote a short note on the system (see below). This elicited a lot of on-line discussion. The system was designed by some of my colleagues at the ANU, but I wasn't involved in development and was using it as an ordinary, but computer literate, voter.

The pilot system is only available at some poling stations (it is not designed for remote use over the Internet). The voter identifies themselves to a poling official in the usual way. Instead of being handed a ballot paper, I was given a coloured card with a bar code on it and asked to select an electronic polling booth.

The electronic polling booths are modified versions of the usual ones: a cardboard box forming a desk, with a cardboard privacy screen around three sides. The box has a hole cut in the top (a rough job apparently done with a knife), through which a computer screen is visible. There is a numeric keypad and a bar code scanner on the desktop. The keypad has cursor arrows and an enter key (all other keys are blanked out).

The voter scans the card they were handed through the bar code reader to begin voting. The polling official warned there had been some problems with the scanners, but I found ignoring their advice and smoothly swiping the card worked well. The voter can select instructions in multiple languages, but I used the default English. A facsimile of the paper ballot paper appears on the screen (with text a bit smaller than is comfortably readable).

The keypad is used to select candidates in order of preference. In my case this required selecting seven candidates (the ACT having an electoral system with multi-member electorates). The cursor keys are used to navigate to a candidate on the screen and then the enter key is pressed to select them as the next preference. The selections are automatically numbered starting from one. I had some difficulty jumping between political parties; the ballot paper has candidates in columns by party and the system wouldn't let me easily move sideways from party to party (having to move up and out of one column and then down into the next).

With seven candidates selected, I indicated I was finished and was asked to swipe the bar code again to confirm my choices. The polling official seemed surprised that I had finished so quickly. The last step in the process is to place the bar coded card in the ballot box (where paper ballots also go). The votes are not recorded on the card, but it is presumably counted to cross check the number of electronic votes.

Clearly a lot of thought has gone into the system design and I found little to fault. It is a reasonable adaption of the paper ballot to an electronic system. In comparison last week I used an automated library book borrowing system, which was a lot more difficult to operate. That was a production (not pilot) system which also uses a bar coded card. The library system was much less user friendly and I was unable to work it successfully, even with help from a librarian.

There were a few areas in which the electronic voting system could be improved:

* TEXT SIZE: The list of candidates on screen was too small to be read comfortably. This was presumably because of the need to retain the same format as the paper ballot and have all names on screen at once. Some way to provide bigger text is needed.

* BOOTH DESIGN: There was the smell of overheating electronics coming from the booth. It would appear that not enough thought had been given to cooling the computer equipment inside. This could be dangerous, as the equipment is housed in a flammable cardboard box and is unlikely to conform with safety standards. This could be worse at non-air conditioned polling booths. A few degrees of tilt in the desktop might have helped the ventilation (as well as the ergonomics), while retaining the appearance and privacy of a conventional booth. It should be reasonably simple for a company such as Datatrax ... which has experience in designing electronic kiosk cases to produce a better booth...

* SELECTIONS: The only way to select candidates was starting from the most preferred to the least. This is the opposite of the process which I actually used. What I did was get the list of candidates from the electoral commission web site and deleted the ones I didn't like, until I had the number of positions available. I then printed this list, took it to the polling booth and transcribed the result to the electronic system...

From Electronic voting in the ACT, Tom Worthington, Link Mailing List, Mon Oct 15 2001 - 14:17:30 EST

Comments in response to this were made on-line.

Several news reports suggested that the electronic system had slowed down the vote counting:

... Internet users following the ACT results have significantly slowed down the collating of electronic votes, ACT Electoral Commissioner Phil Green said last night.

ALP's strong showing puts the heat on Libs By BRENDAN NICHOLSON POLITICAL CORRESPONDENT, The Age, Sunday 21 October 2001

... A trial of electronic voting in the poll ended up actually slowing down the count after the ACT Electoral Commission's Web site was swamped by people following the count.

Labor set for victory in ACT after gain in primary votes By Laura Tingle, The Sydney Morning Herald, October 22, 2001

The election results web site is available. When I tried it on Monday afternoon (2:52pm) it was taking from 2 to 20 seconds to get a page of results, with most less than 10 seconds. This is acceptable for such a system. The suggestion that Internet users slowed down the collating suggests that the database used for counting the results was connected live to the web. This is unlikely, as normally a separate web server would be used to display the results. As well as potentially slowing down the counting system, using it for a web server would increase the risk of a security attack on the counting system.

Remove restriction on pre-poll votes?

Apart from the glitch with the web server, electronic voting was a success. However, to expand the system to all polling places would require many more electronic polling booths. Perhaps next year an expanded pilot could be tried, with the restriction on pre-poll votes removed. At present only persons who are unable to attend a polling place on election day are permitted to vote at a pre-poll voting centre. Removing this restriction would allow the four pre-poll voting centres to collect more votes using the same equipment. Voters might prefer voting at their convenience over a few weeks and, if popular, this might result in one quarter of all votes being cast electronically, with a significant saving in costs.

See also:

Acknowledgement

Thanks to members of the Australian Computer Society, Department of Computer Science ANU and the Link mailing list for assistance.

Further Information:


Comments and corrections to: webmaster@tomw.net.au

Copyright Tom Worthington 2001.