Regulation is a cat-and-mouse game. The mice work hard to find other games when the cat is not away. The same holds for the regulation of elections where cheating has become a game as well as an open secret.
Nonetheless, pundits like Mon Casiple seem to believe in automation as a silver bullet against fraudulent elections in the Philippines. Describing the latest proposal of the Comelec Advisory Council, he states:
Copies of the election return are available at the precinct level (30 copies will be printed!), posted in the precinct, and made available for viewing through internet to accredited parties, including citizen arms and the Kapisanan ng mga Brodkaster ng Pilipinas (KBP). The general public can monitor on real time the national and local count in the internet or through the various media.
In the PCOS system, there are precious few chances for cheaters to intervene directly in the system. Security measures will be in place to thwart intrusions, hacking, or preprogramming of the machines. The more likely time for electoral fraud will be during the period before the voter votes. However, this will be very limited and thus hazardous to the cheaters.”
It is hard to be convinced that cheating is already on the wrong side of history, given the “garbage in-garbage out” nature of automation. Some questions remain dangerously unanswered. For example:
- How can we verify that the number of ballots counted by machine is the same as that of the actual voters? (If we can do this, we can effectively purge the voter list of “excess” voters used in the past for cheating.)
- How do we ensure that in the black box of the computer nothing improper takes place?
- Why restrict internet access to precinct-level data to “accredited parties”? Why not the public?
- Can anyone (i.e. the public) “grab” the precinct-level data and aggregate them independently? Will the accredited parties have their own independent software for aggregating precinct-level data?
The second question above is the most difficult. But we can perhaps deal with the black box and the “newness” of automated counting through some form of redundancy or “parallel” testing. The results of the machine can be compared with that of the traditional count at the precinct. The latter, while imperfect, has generally stood the test of experience with the adversarial counting process that has been open to poll watchers and citizen groups.
For a transition period, the two modes of counting can co-exist, with reasonable rules on which of the two should be accepted. For example, the automated count could be presumed correct if it differed by less than 2 percent from the manual count. The transition also gives the voters and regulators experience on how to improve the automated poll for the next election. This transition is not as expensive as it may seem because candidates and concerned voters already know how to watch the polls, and there is already a Comelec budget for manual counting.
The argument may be made that retaining the manual count simply perpetuates the cheating-prone status quo. This, however, cannot be the case if the main purpose of the manual count at the precinct level is to elicit confidence in the automated system. Moreover, the parallel test does not have to extend to the canvass beyond the precinct. Where there is no further dispute on the precinct-level election returns, the aggregation of the data can be made in the open, available on the web for all to see.
The economic problem in poll automation is that of making the best use of precious public resources. An automated system would be a waste of public funds if it turns out to be badly designed, or if its rules can be circumvented by election cheats. A naïve faith in the advice of technical experts can unhappily result in an ultimate fix that is much more expensive than prevention.
The above has been published as an op-ed piece in the Philippine Daily Tribune, March 25, 2009.
See my earlier post.