Wired editor Chris Anderson has a provocative essay this month arguing that data-mining has gotten so good at identifying correlations that scientists don't even need explanatory models anymore:
The new availability of huge amounts of data, along with the statistical tools to crunch these numbers, offers a whole new way of understanding the world. Correlation supersedes causation, and science can advance even without coherent models, unified theories, or really any mechanistic explanation at all.
Oh geez. Is Anderson saying this just to get a rise out of the rest of us? Does he really think that the ability to reproduce observations counts as understanding?
John Timmer of ArsTechnica has a good critique of this argument. I'd just add that the data-mining examples Anderson offers are themselves based on models (for example, Venter's gene-sequencing system requires knowledge of the basic structure of genes) and that fundamental physics works because the underlying reality of nature has proven to be less complex than what we directly see.
Anderson would have made a stronger case had he stuck to making a methodological point: that data-mining, which used to be pretty dodgy, has become one legitimate mode of doing science, joining the hypothetico-deductive technique and numerical simulations.
--
Edited by gmusser at 06/26/2008 11:59 AM
60-Second Science7/3/09 Genetic Link For Perfect Pitch?Recent research claims to have found some evidence for a genetic link to perfect pitch. Christie Nicholson reports
60-Second Psych6/14/09 Increase Your Creativity: Live AbroadRecent research shows that people who have lived in a foreign country are more creative when it comes to solving problems. Christie Nicholson reports
60-Second Earth7/2/09 The Wonderful World of Transgenic AnimalsFrom spider silk in goat's milk to tomato genes in salmon, genetically modified animals are proliferatingin the lab. David Biello reports