Google Translate

Wednesday, May 03, 2006


Here's a random thought that came to my mind while brushing my teeth:

My favourite author, Isaac Asimov, wrote a short story entitled "Franchise". Set in the future, it was about a man who had been chosen to elect the next President of the United States. This voter had been chosen by a supercomputer, Multivac, to represent the entire population, based on years of voting data and statistics. In other words, he was the only voter -- or represented those voters -- whose vote really mattered in any election.

Truly, one man, one vote.

Then I thought about our current Singapore elections. The country is divided into constituencies and each constituency has several polling districts. Each district has a polling station, which serves the citizens living nearby, usually numbering in the small thousands. It is highly unlikely for a citizen living at one end of the island to vote at a polling station located at the other end.

Therefore, voting in Singapore is localised. And it is based on this localisation that the ruling party is able to determine how groups of citizens voted (though the individual vote is still secret).

It doesn't take a science fiction writer, then, to see that a supercomputer could likewise make use of these localised numbers to predict the outcome at a polling district, then extrapolate it to determine the election result for the country. Given sufficient data on past elections and how each district voted, coupled with demographic and other relevant data, such as wealth distribution, educational levels, current events, social situation, etc., I believe a supercomputer would be able to give a better-than-average accurate prediction.

The only thing it wouldn't be able to determine is free will. I think this plays a much smaller role compared to the other factors, but it is no less vital. That's where the one voter would come in.

After he votes, the supercomputer would go through a series of statements and assign relevant weights, e.g.
"The voter's income is $x a month. 40% of the population earn less than him, 1% earn the same, 59% earn more. He voted at Suahkoo district, which accounts for 1% of the population. He lives with his parents and mother-in-law, similar to 60% of the population. One ex-wife, like 50% of the population. One current wife, like 55% of the population. The USA is at war with Luxembourg. The last time the USA was at war, the vote was such-and-such." etc etc.
(No offence to the USA or Luxembourg.)

To me, it seems that based on all of these data, any reasonable person -- who can absorb and make sense of the numbers -- would be able to make some kind of guesstimate about the outcome of an election. It would then be relatively trivial for a (properly programmed) supercomputer to do likewise. And since a (properly programmed) supercomputer is less likely to cause errors, the prediction would be even more accurate.

Could Asimov's fiction come true?


Technorati tags: , ,

1 comment:

Ed, Edd & Eddy !!! said...

Ah, but I would think that free will can and most likely will overide other factors.

Irrational thinking ia a trait that computer don't have, so someone one could vote based on irrational factors. Like not liking the face of a particular candidate. Or having a personal encounter with a candidate that influences the vote.

Post a Comment