Title: Purposing Writing Assessments: Focusing Complex Constructs in Variable Contexts

Abstract: Writing and language abilities are so multi-faceted and variable that construct models for their assessment are necessarily partial representations of the full construct, designed to fulfill particular purposes of assessment in specific contexts (supported by varying degrees of validation and research). Assessment purposes are either normative (comparing on a common scale all people who take a test, usually for decisions about admission to educational programs, certification of professional abilities, or monitoring an educational system); formative (to inform teaching and learning for individual diagnosis, program selection, guidance, or motivation); or summative (to document and report achievement within educational programs). These purposes overlap and are easily confused in educational practices and policies because institutions want assessments to serve multiple functions and stakeholders. Exemplary writing assessments can be designed to fulfill multiple purposes systematically, as in ETS' CBAL project—assessment of, for, and as learning. But most writing assessments for educational purposes remain limited to certain educational programs, populations, points in the lifespan, languages, genres, and purposes for writing (cf. Hornberger's 2003 model of biliteracy). For these reasons, the design, uses, and evaluation of writing assessments in education should make a fundamental distinction between purposes that either are normative (so should not in principle relate to any particular curricula or teaching) or are formative and summative (which should be based directly on and inform curricula, teaching, and learning). To borrow a metaphor from the design of computer systems, "purposing…reduces the number and complexity of steps required to deploy and configure a data center server" (Microsoft, 2016).