Monday, May 21, 2007

The Road to Automation-Part 1

INK FLAGRANTE DELICTO Many of the women in my family voted on the morning of Election Day last week after which we all met for lunch. When I arrived, I noticed they were casually passing around a bottle with a nice-smelling liquid with which they were cleaning off the "indelible" ink from carefully manicured fingers. Not that they had any plans involving flying and voting, mind you, but being a lifetime afficionado of the fountain pen and inks of all kinds, I was amazed at the remarkable simplicity with which "indelible ink" is managed after voting in a Philippine election. I think the use of indelible ink is no longer an effective security feature, but a social signal that one has exercised the sacred right of suffrage and is available for some chit-chat and unstructured "exit polling". It's purpose is more like Ash Wednesday's gray mark on the forehead, and only a lil more difficult to remove.

For the 2007 Midterm Elections, Comelec assigned over 44 million registered voters to 224,748 separate VOTING PRECINCTS composed of not more than 200 voters each and supervised by a Board of Election Inspectors (BEI) composed of three public school teachers, who supervise the polls, tally the votes and prepare the Precinct Election Return showing the votes received by each candidate.

SEVEN DAYS have now passed since the 2007 midterm elections BELIEVE IT OR NOT:

(1) No one knows precisely how many ballots were actually cast, i.e. what was the TURNOUT?
(2) No one knows precisely how many votes each candidate got and who won, i.e. what was the OUTCOME?

The DATA required to answer both these questions exists, but it is scattered throughout the 224,748 ELECTION RETURNS which were prepared by the BEIs after they tallied up to 200 ballots at their assigned precinct. HANDWRITTEN in words and figures on the ER, with the thumbprints of the BEI members, are the number of valid ballots tallied and the number of votes each candidate received.

Someone still has to READ the words or figures in each ER for each candidate, TYPE it into a computer which can later ADD up the corresponding numbers on ALL 224,748 Election Returns. But who has a copy of ALL the ERs at this point in time?

Comelec Resolution No. 7815 (Jan. 26, 2007) instructs the BEIs to prepare the Precinct Election Return in SEPTUPLICATE (7 copies):
1. The first copy, to the City or Municipal Board of Canvassers;
2. The second copy, to the Commission;
3. The third copy, to the Provincial Board of Canvassers;
4. The fourth copy, to the dominant majority party as determined by the Commission;
5. The fifth copy, to the dominant minority party as determined by the Commission;
6. The sixth copy, to the citizens’ arm authorized by the Commission to conduct an unofficial count; and
7. The seventh copy, to be deposited inside the compartment of the ballot box for valid ballot.
Theoretically therefore, there are four entities who should have copies of ALL the 224,748 Election Returns, or are entitled to them:

(1) The Commission on Elections.
(2) The dominant majority party, Lakas-CMD.
(3) The dominant minority party, Liberal Party.
(4) The Comelec's Citizen Arm, National Movement for Free Elections (Namfrel).

The city/municipal, and provincial Boards of Canvassers receive only the ERs of their constituent precinct ERs, and the seventh copy stays with the ballot box.

THE CHALLENGE: Even assuming we could get physical hold of all 224,748 ERs, the fact remains that the DATA we need is handwritten in words and figures beside each candidate's name on each of 224,748 separate pieces of paper the precinct Election Returns. In this form, the data is said to be in "analog" and not "digital" form. To CANVASS the election now means that a human being has to READ each Election Return and WRITE into an adding machine the number of votes received by each candidate. The analog data must be converted to digital form so that it can easily and accurately be transmitted, processed, stored or otherwise used by a computer.

The Comelec of course conducts the Official Tally via a multistage canvass that first determines the winners in the local city and municipal races, while aggregating and carrying forward the provincial and national results. Provincial Boards of Canvassers likewise determine election winners at their level and aggregate national results from below before forwarding everything to the Congress, which performs the final stage of canvass. This process naturally takes a long time, not only because of monkey business, but because even if everyone were absolutely honest and efficient, the official correction of errors and settlement of disputes and the orderly conduct of a MANUAL election canvass simply takes a lot of time.

Enter Namfrel and the idea of a Quick Count...

CAN THE BIG IDEA BEHIND NAMFREL BE SAVED?

Operation Quick Count
contains Namfrel's latest update on its ongoing count of the 2007 elections. As of 20 May 2007 11:29 pm, Namfrel has tallied 38% of the total or 86,464 out of 224,748 precinct ERS at the counting center at La Salle Green Hills High School in Mandaluyong City, Metro Manila. At this pace, Namfrel is at least two weeks from answering with precision our two questions: how many voters cast ballots and how many votes did each candidate receive?

The glacial pace of the Namfrel "Quick Count" in 2007, does not bode well for its prospects of completion, given what happened in 2004 and reported problems already this year with software and human error.

"Namfrel must die!" TONY LOPEZ of the Manila Times declares -- fuming that the Namfrel QuickCount™ has fizzled out in 2007 just like it did in 2004:(via MLQ3)
Whatever Namfrel’s reason for existence, it should have evaporated by yesterday. The much-ballyhooed Namfrel Quick Count fizzled out. Just like in 2004, Namfrel failed in its mission, dismally, which is inexcusable considering that the self-proclaimed poll watchdog is run by do-gooding Catholics and businessmen inured in the art of management and handling difficult logistics. It is manned by what is claimed to be half a million volunteers and equipped with hundreds of computers.
With Chairman Ben Abalos vowing to finish the official tally in TEN DAYS, and Namfrel reporting mysterious problems with its software, it is not inconceivable that Comelec could finish before Namfrel. IF that were to happen then Namfrel would indeed be DEAD and Tony Lopez's imperative would be fulfilled.

But if the Board of Election Inspectors can typically count the votes and produce their ER within 24 hours of polls closing, why does it take both the Namfrel several weeks to add up the numbers that are already written out, in words and figures on the ERs?

Why can neither the Comelec nor Namfrel, even today, give us an exact figure for HOW MANY VOTERS cast ballots last week?

The simple reason is this. It is a very, very hard thing to actually bring together in one tall pile, all 224,748 Election Returns! I wager that NO ONE, not Comelec, not Namfrel, not Lakas, not the LP, will ever actually have every single ER they are entitled to, all in one place, where a dedicated staff might get to work on them.

At La Salle Green Hills High School, at the Namfrel CountingCenter, one finds the closest thing to an attempt at quickly counting those ERs that the Namfrel's half million volunteers have managed to fax, email, text, or otherwise upload to the Center, from the archipelago's far flung precincts.

Comelec's multistage canvass, which may be slow, is actually more logically organized in some respects. Consider a physical fact. The average distance that an ER Sixth Copy must travel to get to the Namfrel Counting Center in Metro Manila is far longer than the distance it will travel to municipal, provincial and Comelec counting centers. As late as last Friday, Namfrel reps were admitting on tv that they do not yet have all the ERs, many being enroute or being transmitted, audited, and checked by overworked volunteers.

It is not surprising to me that neither the dominant Majority nor Minority party mounts its own Quick Count, despite being entitled to official copies of all the Election Returns. The logistics, organization and discipline required to do it properly would be daunting, and not even the Comelec has achieved it, nor does it seem inclined to do so.

But from a purely technical standpoint, an important lesson or insight can be gleaned:

The most logical place to AUTOMATE the Philippine Election is at the Precinct Level, when 750,000 public school teachers are actually available to DIGITIZE the Election Return.

16 comments:

Jon Mariano said...

Why would the Comelec ever want to count the votes faster? There's just no reason for them to want to. More so NAMFREL, they will become irrelevant once it happens! It's better for these two entities (Comelec and Namfrel) if the counting system remains the same.

mlq3 said...

but what happened to our consensus that up to the precinct level, our elections are ok? that if there was a way to aggregate the precinct results without the intervening municipal and provincial stages, the results would be more honest?

MBW said...

DJB,

On election automation.

As you know, France just had its presidential election but for the majority of the precincts, automated election, they were not.

Majority of voting was done manually. Matter of fact, equipment was done away with during the runoff (May 6) after the fiasco in many of them caused delays in getting voters to use them in some precincts in France. According to some election commissioners, some precincts took out the voters' machines in the middle of the voting day on April 22 (1st round) and simply put back the old, tested method of manual voting.

Even then, by 8 PM, when the last vote was counted at the last minute of the last hour, the new president's name was announced right on schedule.

The election automation machines in France are supposed to be top of the range but it's not the automation itself that's the problem (although I believe they are too complicated), it's the educating the voters on the use of the machine.

I voted the old fashion way - manual (putting in the ballot containing the name of Sarkozy in the envelope which I then inserted in a transparent ballot box after my elector's card was checked, cross-checked and my signature verified, etc.)

Same old fashion manual voting will be in place in the forthcoming legislative elections here. (And we are talking here of 86% of the more 44 million French voters who cast their ballots.)

I guess it's because we vote on party listing in France so ballots are already filled up with the names of the party candidates; there's no need to write them; you simply insert your choice of ballot in the envelop and then cast that ballot in the box (except for the presidential election, you choose ONE NAME; party affiliation does not appear on the ballot).

All said, it's still all a question of cheating-less election procedures I suppose.

Amadeo said...

In any calculated move toward automation, beyond just the speed and efficiency that will hopefully be acquired, I believe one also has to determine what would be considered allowable or tolerable “human intervention” that is necessarily incorporated in the total process, given the specific realities where the system will be applied.

Relying on current technologies, one still cannot discount the critical role of the human element/intervention, especially in the possibilities to compromise the vulnerabilities of any system.

As they say, the machine is only as good or as bad as the people using or running it.

Thus in more mature and developed democracies, one would believe that a high degree of the human element is tolerated because of the level of maturity and/or experience of the electorate.

Willy B Prilles, Jr said...

Taking DJB's cue, I'll probably do it by exploring the use of cameraphones to take a snapshot of the ER and send them via MMS to a central node for the quickcount.

A 2-megapixel cameraphone will probably do the trick. It's still pricey now, but in three years time, that kind of cellphone will be pretty much entry level. At 224,748 ERs and 5T per unit, that will amount to an additional investment of P1.1B. Teachers should get even to keep that cellphone for all their work.

Deany Bocobo said...

MLQ3,
It is precisely because the election process is most transparent at the Precinct level that that is the best place to digitize the data. Once the data is digitized and published, it is most secure.

But the reasoning in this post is logistical. While the BEI is at it preparing the ERs, they may as well perform the crucial task of analog to digital conversion.

Willy B Prilles, Jr said...

Just to give you an idea of the timeline involved, in the precinct I observed, counting of the 160 votes cast took something like 7 hours (from 4-11 pm), while filling up the forms, including the ERs took another 3 hours (from 11pm-2am).

Analog-to-digital conversion should, I think, not add to that burden as much as possible; hence the cameraphone idea.

Or, if the manual tallying by the 3rd member can be done electronically, then when the count is over -- at around 11 pm -- the data would be about ready for transmission.

Deany Bocobo said...

Willy a hi-res photo of the ER would serve as a good RECORD of what it contained. But for the purposes of a Quick Count, even if the photographs were sent by MMS to a central location, there would still remain the problem of having to READ the data off the photograph, and encode that into a computer.

What I think I am suggesting is that if the BEI spent most of its time ENCODING THE ER and then transmitting the Digital ER via cellphone...then you'd be getting dangerously close to "automation".

Thanks for this comment...it segues right into Part 2 because I want to address what "automation" is actually all about in the next installment...

Willy B Prilles, Jr said...

Ok, DJB. I'll look forward to it.:)

But allow me to think ahead, on the basis of your premise.

I was actually reconsidering giving the encoding task to the poll clerk -- the 2nd member of the BEI -- who, aside from preparing the minutes, actually was the one who filled up the ER, which the two others signed afterwards.

But then again, in a low-trust society of ours, her task does not allow the poll watchers to really see and check if the data he/she is putting into the ER is correct.

But if a cellphone is going to be used to transmit the data, then what I will probably do is (1) for at least two SMS templates containing (a) the senatorial candidates and (b) the partylist contenders to be prepared by Comelec Manila; (2) for these to be made available to all BEIs; (3) for the BEI chair to plug in all totals for every candidate into the template, using the analog ER as reference; (4) for the BEI chair to initially send the draft SMS message to majority and minority party reps for verification; and (5) once verified to be correct, for the SMS message to be transmitted to the central node.

Deany Bocobo said...

Willy, thanks for chasing the rabbit down the hole with me.

Regarding verification, I think the idea is that once the digital ER is encoded and ready for transmission, WHERE should it be sent?

I think it should be sent to Comelec HQ of course where it will be PUBLISHED in a database website that anyone is allowed to interrogate with an Internet access. Including anyone at the precinct who may not trust the poll clerk.

The publishe version of the ER becomes the ONLY copy of it, and the official one. I think some provision can be made for corrections due to challenges and disputes later on in the canvass, but initially, before the BEI sends it, the public and the watchers may be invited to inspect its contents before transmission, and then after that on the main website.

So the concept behind securing the data is to PUBLISH it where everybody can see it 24/7.

BTW, the cellphone that is included with the ballot box can have special programming that makes al this easier, including transmission, which can be made a lot easier than sending a bunch of sms messages.

It can also have a global positioning system as an added security feature by sending its location coordinates when the transmission is made. Talk about PRECINCT MAPPING this is gonna be high tech!

The key step can be taken by the BEI though: DIGITIZE the ER!

the jester-in-exile said...

on the automation of polls, djb, perhaps the POV of this engineer might be of some value, written early this year: part 1 and part 2.

my two cents. i've always believed that automation for speed is half the battle; the other (and more important) half is identity security.

Willy B Prilles, Jr said...

DJB, I will add the citizen's arm rep at the precinct level, whoever it is going to be, to the dominant majority and minority party reps in the verification process, consistent with the provision organizing the BEI.

If all four parties agree (the BEI, the two parties and the citizen watchdog), there will be a higher degree of certainty on data integrity.

Because these are for national positions, the only logical destination would be Comelec Manila which should make this available 24/7, with automatic real time updating capabilities. (Think of the NFL or NBA web-based scoreboards which automatically updates every 30 seconds or so.)

If this design is proven viable, then there is no reason for coming up with separate templates for local positions (provincial and city/municipal), with data being plugged in by the BEI, verified by the three other players at the precinct, and then transmitted to BOTH the provincial and municipal node, which should have the same 24/7 website.

Bokyo said...

Hi Dean,
I wish our election process will be like this.
1) After the precinct canvass has been finalized, a report (talking about computerized here) will be prepared also (besides the ER) enumerating how each voter voted which has been identified only by number (known by the voter only and a certain list) instead of names. This will then be posted in a bulletin board in the precinct. This way each voter can check how or if their votes have been counted. This is a reverse tally.
2) Same will be done to municipal canvass how the precinct totals have been posted and added up, then also to provincial level which should also be posted back if possible in all precincts.
An election as clear as this will erase all doubts and can easily be verified. Anyway it's only a wish. If only the COMELEC is very efficient and truthful to their jobs we would not be thinking all about this.

Bren said...

Maybe the process fails --- the process of counting the votes with integrity --- because Filipinos simply do not want to pay the price to make the process work well.

Without the will, the way is blocked.

engineerOFW said...

Better than digitizing the ER is to digitize the individual vote.
The individual data-record to contain
-- the actual votes
-- date/time-stamp (to the second);
-- precinct-location
-- security check-digit-01
-- security check-digit-02

john marzan said...

Many of the women in my family voted on the morning of Election Day last week after which we all met for lunch.

were you able to vote, djb? sino binoto mo sa senado?