I read with great interest the recent article in the Atlantic about
large-scale data and the advent of a new science based on predictions
from complex network models. The implication was that we are not
going to have intelligible theories in the future but rather we will
obtain our predictions from black boxes. This is somewhat
unsatisfying. I think it is possible to have complex models that
cannot be written down on a piece of paper, yet to abstract from them
some clear principles, which are not true all of the time but which
still allow people to discuss and reason about the models. A key
aspect of this reasoning is that people can combine different black
boxes or modify them only if they have some understanding of them.
Abstracting the essence of the problem is very fundamental to
scientific theory and I do not think should be lost in this new
science of networks.
Unpublished letter in response to:
DAVID WEINBERGER - To Know, but Not Understand: David Weinberger on
Science and Big Data
http://www.theatlantic.com/technology/archive/2012/01/to-know-but-not-understand-david-weinberger-on-science-and-big-data/250820
