Guts and Brains

Nice article in the Times today about the BCS system for computing college bowl matchups. The writer created an essay-grading algorithm that agrees with humans more than they agree with each other (fits in well with the Gender Genie, see earlier entries). He says, among other things:

1. Playoffs aren't the best determinant of a team's ability -- the best way is to load the stats into the computer and have it play hundreds of simulated games
2. Even so, more work can be done to bring the sportswriter's picks in line with the computer picks

The thing to do is change the computer weighting of 'difficulty of schedule', 'size of the win' etc...

I can use this perspective for my ongoing approach to health stats. Dr. Pf, Ralph Nader and probably countless others continue to marvel that our society freaks out about SARS, about West Nile, and (though it's not the same) Viet Nam or 9/11, yet does very little about the number of traffic deaths each year, the number of people dying of lung cancer, the number of kids who fall out of windows...

One upcoming post should be all stats from the Bill Bryson book I'm reading (especially the numbers on fossilization rates). But right now just remember that 300 million people died of smallpox in the 20th century, more than all the wars and genocides and state-sponsored murders combined (which I think totals 100-150 million, depending on China totals).

Yet the eradication of Smallpox was not celebrated with the gusto of V-E day, say. And of course it couldn't be: Stopping a virus doesn't give the emotional satisfaction of stopping an army. Look at the Iraqis who didn't feel free until Saddam was physically captured, six months after it was clear he could never rule again.

But we have to work toward intellect and evidence-based decision making, over emotional decisions. In medicine it's still a problem -- there are plenty of useless antibiotics prescribed
just to 'do something', plenty of CT scans ordered 'just to see' when in fact there's no indication of anything to see.

But more importantly, as seen in my recent medical errors seminar, it seems like we allow 'a plane crash a day' of deaths due to communication errors, transcription errors, or missed diagnoses. We'd never allow a plane crash a day -- or a month, even -- from the aviation industry.

From this, we must deduce that the emotional impact of a plane crash hurts everyone -- families, airliners, economies, governments -- much more than the impact of diffuse medical errors, which, I guess, primarily hurt the families, but also the doctors, hospitals, insurance industries...

If we weighted the intellectual response to deaths more than the emotional response, we'd overhaul the healthcare delivery system overnight, and maybe lighten up on airline safety standards. We'd stop reporting on the flu outbreak and start reporting on traffic safety and CAD.

What attracts me to this line of thinking is you can apply the BCS-type algorithms to predict the public's (and media's) response. I touched on this in my Station Fire piece in the Herald, with the local vs. foreign angle, and rock fans vs. hip hop fans.


But the biggest factors in reporting deaths seem to be 'can it affect me even if I act responsibly?' (communicibility of SARS, flu, even though falling out of windows or traffic accidents are more widespread) and 'is it the result of cold-blooded intent' (9/11 killed 3000 people, possibly less than 10% of the Iran earthquake deaths (a sudden calamity), and a tiny fraction of those who will die of medical errors this year (a widespread, diffuse calamity)).

I suppose it will be harder for algorithms to predict public awareness of celebrity deaths, the shuttle disaster, etc. Aaron Schatz on the Lycos 50 reports on this regularly, and has some thoughts on which celebrities get the most searches when they die...

Finally, the way to measure public response to tragedy is as simple as counting newspaper column-inches, or minutes in a newscast... or as complicated as looking at new legislation, new money devoted to the problem, new policy changes.

I'd like to think, ultimately, we'll save more lives when, make people happier, etc, when we move toward appropriate quantification and response to these deaths... but there is something to be said for the emotional response. The most obvious example: Maybe the US prevented more 9/11's by going full-tilt against Al Qaida and Iraq, instead of shrugging it off. Maybe the muted response to medical errors or viruses, as opposed to plane crashes, or mass murders, is our society's way of trying to cope with things that can't be changed, and address things that can be... But everything can be changed, if we agree on it and make it a priority.