In Defense of an Instrument-Based Approach to Validity

摘要

Paul E. Newton (this issue) argues in favor of a conception of validity, viz, "the consensus definition of validity," according to which the extension of the predicate "is valid" is a subset of "assessment-based decision-making procedure[s], which [are] underwritten by an argument that the assessment procedure can be used to measure the attribute entailed by that decision," (Newton, this issue). Standing in contrast to this conception of validity is Borsboom's (2005) account, according to which the extension of the predicate "is valid" is a subset of psychological tests. In Borsboom's own words, "a test is valid for measuring an attribute if and only if (a) the attribute exists and (b) variations in the attribute causally produce variations in the outcomes of the measurement procedure," (Borsboom, 2005, p. 150).' The consensus definition of validity and Borsboom's instrument-based accounts are clearly incompatible, and in the course of defending the former, Newton argues against the latter in an effort to motivate his own account. In what follows I assess the two principal reasons Newton gives for rejecting Borsboom's analysis. The first pertains to differential measurement quality in light of four considerations: adherence to proper measurement procedures and guidelines, the context of the measurement, characteristics of the group being assessed, and the use of measurement outcomes. The second objection is, in effect, the charge that Borsboom's account (and the instrument-based account in general) underrepresents the concept of validity. I will argue that Newton's objections to the instrument-based account either constitute no serious objection to the instrument-based account or they are question begging.