Monday September 23, 2013
KL 232 (Chancellor’s Conference Room)
Factor analysis: Underdetermination and uncertainties
The underdetermination of scientific theories by their evidence is a longstanding issue in the philosophy of science. This matter is prominent in psychology, which frequently employs abstract theoretical entities. This paper considers underdetermination in the factor analytic model (FA), which is often used to bridge the gap from concrete data (e.g. test scores) to hypothetical constructs (e.g. intelligence). After introducing these issues, three general topics are addressed. (i) Underdetermination: FA presents a philosophically interesting (“real-life”) form of underdetermination, illuminating and/or conflicting with several claims about underdetermination, abduction, and theoretical terms. (ii) Uncertainties: FA helps distinguish at least four kinds of uncertainties. The prevailing practice, often encoded in statistical software, is either to simply ignore the most difficult kinds, or to lump them in with the most tractable kinds. (iii) What to do: some suggestions for dealing with these hardest types of uncertainty are offered.
Kent Johnson is an Associate Professor in Logic and Philosophy of Science and the Institute for Mathematical Behavioral Sciences at UC Irvine. He received his Ph.D in philosophy and certificate in Cognitive Science from Rutgers. He works in the areas of philosophy of linguistics, philosophy of psychology, cognitive science, and philosophy of language. For more on Johnson’s work, see http://www.lps.uci.edu/~johnsonk/