Kuhn’s basic idea starts from the premise that all scientific models are imperfect approximations of the truth. My take on this: Much like Gödel says that any consistent system of axioms in mathematics is inherently incomplete, science will always have some observations that it cannot account for. In statistics, for example, the ideal formula for a line y=mx+b becomes y=bx+a+ε, where ε is the error term, i.e., all those things we didn’t measure that also effect the relationship between y and x. These unaccounted-for real-world influences cause the data pattern to be something like a cigar-shaped scatterplot rather than a perfectly straight line. We can try to get closer to a straight line by measuring more things—y=b1x1+b2x2+b3x3+a+ε—but while we can reduce ε, in practice it never really goes away.
|This is the edition of Kuhn that I have.|
|I couldn't resist.|