I enjoyed the May 11 article on Swanson and Smalheiser's ARROWSMITH program for finding hidden connections in MEDLINE. I also found the cited opinions of critics and skeptics of their work to be specious.
1. The claim that "everyone has some skepticism that you can find something new from what's already out there" in the medical literature is curious, since the world is certainly "out there" and scientists find new things all the time. Diseases, elementary particles, and chemical reactions are just as out there as MEDLINE; a main obstacle is simply making the right connections, and ARROWSMITH helps accomplish this using simple and clever methods.
2. "What if the migraine headache was caused by something we haven't even looked at yet? In that case, no matter how hard the machine flails away at this, it's not going to find an answer." Sure, ARROWSMITH isn't infallible. Substitute "scientist" for "machine" and notice the double standard.
3. Even if ARROWSMITH fails to suggest interesting discoveries in the humanities, so what? Many medical researchers "fail" in the same way.
4. The demands put on the user in terms of both the time investment to learn the program and the expertise required are significant, but so are the demands occasioned by any new scientific instrument. I doubt ARROWSMITH is more taxing than, say, brain imaging.
In another section the real issue is too briefly mentioned: allowing computers to generate hypotheses offends many sensibilities. In the book "The Same and not the Same", the Nobel Laureate chemist Roald Hoffmann, in his discussion of programs that are able to plan organic syntheses, remarks that
The psychology of human beings is not well suited to admitting that we can be replaced by a computer program, only that others can be.
This phenomenon is seen everywhere where such programs have been developed: the GRAFFITI program that generates conjectures in graph theory developed at the University of Houston, and our MECHEM program that generates reaction mechanisms in chemistry, to name just two.
People and scientists should build on the new capabilities that are offered by computer programs rather than hurl specious criticisms.
Sincerely,
Raul Valdes-Perez