Election Epistemology

The Bush team is arguing that a manual count of votes is less reliable than a machine count. The Gore team is arguing that a manual count is more reliable.

How can anyone possibly tell which is right? You could run a manual count, but then what would you compare the results to? Your own personal manual count of 5,700,000 votes? A machine count? Whatever you're comparing the result to, you're saying that that is the most reliable. But how do you determine that your standard is the most reliable method?

Maybe you could run each type of count many times, and see which has the narrowest spread of results around its average. But that doesn't prove anything. A machine count has a very narrow spread (virtually the same results every time), but it makes systematic errors. Also, one well-placed computer hacker could swing an election by 10% and probably no one would be the wiser. Computers are dangerous.

Probably the manual count is more immune to a small number of people causing a large distortion. But the manual count is still prone to systematic errors due to quirks of human perception and cognition, and to the layout of the ballots. The error could easily favor one candidate over another.

So all the (reasonable) methods seem prone to different kinds of errors, and there's no way to compare them numerically so that one of them would have the "least amount of error". Hence any candidate's attempt to argue that a particular method is "more accurate" is really just an attempt to get people to choose a type of error that favors that candidate.

Conclusion: No decision about whether or how to recount the votes in Florida could possibly be fair? (Same with re-voting.)

Any thoughts?

-- BenKovitz

Do both. In fact, let any interested party which can demonstrate reasonable competence (eg the parties, the feds, the ACLU, the UN) make its own count, however it likes. Recount until there is a 95% confidence in the outcome (computed with a t-test or similar).

CategoryVoting