ABSTRACT
An inevitable feature of item generation, as considered here, is the delivery of items that have not been trialed. Another is the inability to make perfect predictions of item response characteristics for these items. Successful implementation of item generation requires a defensible methodological framework for dealing with these challenges.
