ABSTRACT
A far more productive approach is to design items with foresight that keeps Automated scoring (AS) in mind from the outset, so that the items might be better positioned to both address the construct as intended and to be readily amenable to AS. Designing a system for such flexibility in demand and potential changes to testing programs in the future creates further complexity in producing a robust and functional scoring process. The overview of human scoring sets the stage for what an operational AS system must also be able to accomplish if it is to match or improve upon human scoring. To be comparable to this human system it would have to have the same level of concern for ensuring the quality of scores in managing responses and potentially offer notable improvements in managing raters since the number of raters could be reduced, potentially dramatically, by a high-performing AS system.
