Just to prove that operations research is relevant everywhere, Ed Kaplan of Yale University has pointed out a fundamental flaw in the voting process for American Idol. It appears that lots of people get busy signals when trying to vote for “their” idol. This may appear to be simply a frustration but actually gets to the heart of the fairness of the voting system. As Ed argues in an op-ed at the Hartford Courant, if votes are limited due to phone capacity, and each candidate has a separate line (each with the same capacity), then the votes will by necessity be much more evenly spread than the electorate. So a runaway winner can turn into a coin-toss.
In the best OR tradition, Ed not only identifies the problem but solves it:
There is a simple solution to this problem that does not require “American Idol” to install additional phone capacity. Instead of assigning each contestant a personal phone number, use a single number for all voting, and have voters select their favorite by pushing a button after the call has gone through. This simple fix would equalize the chance of encountering a busy signal for all callers.
Many votes would still be lost. In fact, if the total phone capacity was left unchanged, more calls would be lost due to the increase in processing time per call necessitated by button-pushing. But, since all calls would have the same chance of getting through, the total votes received would be a representative sample of votes cast.
If only OR people ran the world, even American Idol would run better.
> “If only OR people ran the world, even American Idol would run better.”
Ha! Maybe you mean “economists”. 🙂 They know about preferences and incentives.
In this case, I mean OR! The issue is not preferences or incentives. It is constraints and real-world operations. And that is what OR is good at.
I don’t know about American Idol, but good telephone voting systems use the same back-bone for all the incoming numbers. That essentially puts the button-press at the dial, solving your problem.
At least, the is the way these kind of systems work in Sweden.
But if the backbone is not a source of congestion, doesn’t the problem remain? Or is the backbone always the source of congestion in such systems?
The dialidol.com data clearly report quite different likelihoods of encountering a busy signal for the different contestants, and these are sampled every 3 minutes! It could be that there is a common congestion block on local phone exchanges, but overall it is quite clear that people trying to vote for different contestants are experiencing quite different probabilities of encountering a busy signal, which greatly distorts the results.
I doubt that there would be something like American Idol in that case!
As far as I understand it (the issue has come up several times in Swedish media after various phone-vote shows), the backbone will, in case of congestion, sample more or less uniformly from the incoming calls.