Computers cannot heal

As computers appeared on the public scene in the ’50s, psychologists marveled at the parallels between data-processing technology and the workings of the human mind. A handful of scientists of the day believed that the functional structure of the human mind was fully analogous to that of the computer. Considering that computers in the ’50s were slow, noisy, and unreliable, those scientists apparently didn’t think much of their warm-blooded subjects. I happen to be old-fashioned enough to think that some things will always be beyond man’s understanding, with or without a supercomputer in every garage.

Indeed, I believe computers are far more ignorant about the way humans think and behave than most people suspect. Medicine, for example, is often regarded as a field in which computing and humanity have intersected to create an era of unprecedented advancement. Patients get irritated when tests take days or weeks to come back from the lab. Does it really take that long to feed something to a machine connected to a computer and print a one-paragraph report? That’s the image we have of medical lab work, but the heavy lifting isn’t automated. We can’t feed a tissue sample to a computer for a diagnosis, but lab technicians can consult computers, as well as printed texts filled with illustrations and analyses that computers helped to assemble.

The reality is that, except for recording results or simplifying the presentation of visual tests, computers aren’t a whole lot of help in diagnosing the ills of their creators.

Computers are consistently good at one thing that’s of limitless use to medicine: control. A computer can put what you want in precisely the right place exactly when you want it to be there. A computer can precisely control the strength, position, and duration of a beam of laser light or gamma radiation, making possible modes of therapy that weren’t possible or practical, under human control.

Control is where we should focus our attention. Don’t worry so much about teaching computers what’s going on inside our bodies or our heads; we may just waste effort encoding knowledge that we already possess and that we effectively pass on through ordinary education. Computer-controlled apparatuses are of limited use on humans now, mainly because computers are imperfect and inefficient at sensing.

Computers will eventually see better than humans. Perhaps they’ll take some of the work out of reading films from X-ray, MRI, and CT scanners. I hope they’ll gain the ability to feel, not as humans do, but enough to automate and lower the risk of touchy procedures such as the placement of spinal catheters; a machine could adjust if the patient moves or if it encounters an unusual amount of resistance from the tissue. If computers can learn where and how much we hurt -- leaving it to humans to understand why -- they can dispense anesthesia or electrical stimulation where needed, and at the correct dosage.

The question facing those who work in the sciences is whether research into computers’ involvement with medicine should proceed with the expectation that there are no limits, or whether it should be done within limits that experts consider feasible. I wonder whether computers can help us answer that dilemma.