I helped to organise the last Ruby Manor conference (http://rubymanor.org/3). As an experiment, we built a web app (http://vestibule.rubymanor.org/) which allowed anyone to submit, discuss and vote for proposals. When voting closed, the 8 most popular proposals were automatically selected.
Is it brash to suggest that maybe those "8 white guys" actually did have the most merit out of all candidate speakers, and that you would have done the conference a disservice by bumping one of them for someone else, based on sex or gender alone?
Whatever racial or gender or economic biases might exist that influence the shape of the global developer community, these need to be addressed at the source: in kids' youth when many programmers take on computers as their lifelong obsession, or in high school curricula and college programs.
The conference organizer's job, as it pertains to race and gender bias, is to provide a neutral selection process that chooses the best speakers on their own merits. Their job is not to diminish the quality of the conference by imposing their own, nobly intended but horribly misguided, form of bias on the candidates. More than hurting just the conference or the candidates, I fear that this can have serious unintended consequences relating to public perception of the quality of e.g. female programmers, in the same way that "affirmative action" does on college campuses.
I think a simple estimate of the composition of the "most merit of all candidate speakers" could be to compile statistics of the contributors to the 1000 most popular gems and rails-core, or something similar. There's more heat than light in this controversy, and for a discipline that prioritizes CS techniques as a proxy for skill the flashlight is readily at hand.
While I appreciate that you're looking for solutions instead of trying to amplify the finger-pointing, the approach you are imagining is not going to generate a very good speaker line-up.
There are a huge number of variables to consider when planning a speaker line-up. As I said before, the best conferences are carefully crafted around a theme or issue. Even in the case of a regional conference, it doesn't require that the speakers live nearby. Just selecting popular, smart or friendly devs is not a theme unless your theme is "Random". :)
Sometimes there's a milestone library or development approach that you want to shine a light on. Perhaps 5 of the top 10 gems are template engines or test frameworks, and in 2012 that's a dull theme. Factually, some developers are great speakers, but many are not able or willing to give a good talk. Some speakers are amazing but have nothing interesting to show off or discuss.
Some speakers live far away or have special needs. You might be able to afford and accomodate one or three, but a roster of expensive tickets would make the event lose money.
However, the biggest reason that the algorithmic curation approach is not going to work is that curation is less like selling seats on a plane and more like making a mix tape for a fickle lover. You need to delight, surprise and enrapture people. A month after the event, they remember not what was said but how they came away feeling. It's hard but worth it to find ways to inject ideas that people are not expecting.
It wasn't meant to generate a line up, only to provide some idea of demographics. You have to start somewhere, even if only to figure out how true "curating on merit gives you white guys" is.
The result? 8 white guys giving talks (http://rubymanor.org/3/videos).
We’re going to try something different next time, but who knows if that’ll work any better.