Long before the health care industry became so enamored of electronic medical records, researchers were struggling with the problems of how to enable a computer to translate languages. Over the years, computer translations have become much better than they were, and now anyone can generate a web page in a different language quite easily.
The issue, however, is how accurate a computer-generated translation can be. In the Slate article by Konstantin Kakaes, “Why Computers Still Can’t Translate Languages Automatically,” the subtitle says “We need to teach machines to understand the meaning of words.”
This is an automatic punt back to the rule-based notion of human thinking and human expression. Herbert and Stuart Dreyfus again. Novice vs. Expert again. The compulsion to believe that anything can be turned into a series of on and off switches.
With computer translations, Kakaes tells us:
The central questions guiding most of these [translation] projects is: How can you tell when a translation is any good? Even humans struggle to rank different translations. This makes the challenge of automating evaluations ever starker. And if you don’t know or can’t assess how well you’re doing, it’s hard to improve.
Amen.
For starters, let’s get on the table that good physicians practice the art of medicine, not cookbook medicine, and the nature of the expert knowledge practitioners offer their patients is often not reducible to rules.
Don’t get me wrong. I am not against computerized check lists suggested by Atul Gawande in The Checklist Manifesto: How to Get Things Right. No one is arguing that this kind of attention to detail decreases medical mistakes. Nor am I against electronic medical records. What I AM against is the implicit notion that everything a health care practitioner does can be reduced to the choice between yes and no.
Our health care system is in serious trouble today because regulators live by this notion. In the practice of medicine, this yes-no decision tree can be very helpful. Certainly, as Gawande says, checklists can reduce medical errors. But make no mistake about it. This decision tree isn’t the practice of medicine.
In the flurry to create rule-based protocols which become the practice of medicine, we have created quality control initiatives, standards of care, checklists, protocols, and evidence-based medicine. And to this list should be added the massive current effort to convert paper to electronic data in the health care world. It would behoove us to keep in mind what computers and yes-no decision trees can and cannot do.
We are currently mistaking the yes-no rules for the practice of medicine. There will be no real health care reform until those attempting to bring about changes in health care understand the difference.
