What's This is not to be confused with the original and star, Bob Park's What's New.

1. The Obsolete Scientific Method:
   The editor in chief of Wired, Chris Anderson, has written an article that claims the scientific method is becoming obsolete as researchers develop the ability to collect, store and process petabytes of information. You see, the only reason we needed the scientific method and its slow procession from hypothesis to testing to refinement to repeating, is that we have never been able to collect a truly complete data set! I am sure plenty of people will point out the absurdity of this claim; two of my favorite sites already have. I humbly add today's edition of What's This to the growing collection of such criticism.
2. But What Will Happen Next Time?
   Anderson suggests that with enough data it is possible (nay, required) to “throw the numbers into the biggest computing clusters the world has ever seen and let statistical algorithms find patterns where science cannot.” Then, we simply review these computed correlations and enjoy the scientific progress. We don't need models when we have a complete data set. One problem, however, is that we can never collect data from the future. Models are used to provide predictive capability. A petabyte of data regarding the motion of the planets, even the plutoids, will surely allow an “analyzer” to describe their motion to incredible detail and precision. Where will these objects be in one year? What about 100 years, 1,000 years? An analyzer doesn't care about the future because their job is to analyze up to the present, which is why I didn't call that person a scientist. Assuming that the behavior up until now accurately predicts that of the future is to also assume that the system exhibits no non-linear behavior. Reviewing the behavior of some chaotic systems will demonstrate that measurements of the past and present cannot always lead to accurate predictions of the future.
3. Mainstream Contradiction:
   Anderson cites the Cluster Exploratory supercomputer as an example of how this new thinking about science is going “mainstream.” It will have a lot of processors and a lot of hard-drive space so it must be a completely new paradigm of performing research. He then notes that an early project of the machine will be to simulate the brain. Simulations are computer implementations of models! Supercomputers do not typically analyze data (though I would be happy to receive correspondence of counter-examples). We only collect as much data as needed to vest a model. Models are continuous while data, regardless of the resolution or total amount, are discrete. Not even Google, the idolized soothsayer of Anderson, can collect so much data that models and the scientific method become obsolete.
4. Maybe He's Right:
   I am forced to admit a successful application of Anderson's highlighted method. My research question is: “Is society developing a heightened inability to understand modern science?” The old way to proceed with this research issue is to construct some models of society and the act of understanding and then to begin fitting data points (people) to this model. But why cling to these old ways! My usage of the new science allows me to let the cloud present a bunch of examples that allow me to conclude: “Yes, a lot of people do not understand how science operates.” If I can get a project approved for the Cluster Exploratory, then maybe I can harness the 1,600 processors to scour the web and find a petabytes worth of examples, thereby proving my argument conclusively.

The opinions expressed in this article are not those of a Ph.D. physicist, but they will be.

Discuss this Item (opens forum in a new window, registration is not required)

Technorati Tags: wired, scientific method, media criticism